The leaf PR. Replaces the placeholder SKILL.md from PR 1 with the real
5-step workflow that loads the per-topic references on demand
(skill-creator's progressive-disclosure pattern), and adds a top-level
orchestrator that runs every tier and reports a pass/fail summary.
SKILL.md changes:
- Frontmatter unchanged from PR 1 (already covers the trigger phrases
and out-of-scope cases)
- Body rewritten as a 5-step workflow:
1. Lint (load escape-hatch.md if blockers)
2. Plan (load api-map.md, then per-topic references on demand)
3. Generate (HF index.html with paused GSAP timeline)
4. Validate (render_diff.sh against per-tier threshold)
5. Document gaps (TRANSLATION_NOTES.md if needed)
- Includes a "Source contains -> Load reference" table so the agent
only loads the references the source actually needs
- Documents the validated baseline numbers (T1 0.974, T2 0.985,
T3 0.953, T4 8/8) so reviewers can reproduce
- Calls out the critical Remotion encoder config (PNG + BT.709) that
avoids the ~0.05 SSIM hit from yuvj420p vs yuv420p
Orchestrator (assets/test-corpus/run.sh):
- Iterates tier-1-* through tier-4-* directories
- T1-T3: setup -> lint -> npm install (lazy) -> render Remotion ->
render HF -> SSIM diff at the fixture's expected threshold ->
generate strip on failure
- T4: validate.sh (lint-only)
- Emits run-report.json with per-tier pass/fail and aggregate counts
- Accepts a single-tier argument for fast iteration: ./run.sh tier-1-title-card
Validated end-to-end on a clean checkout:
▶ tier-1-title-card → mean SSIM 0.9739 (≥ 0.95) ✓
▶ tier-2-multi-scene → mean SSIM 0.985292 (≥ 0.95) ✓
▶ tier-3-data-driven → mean SSIM 0.952941 (≥ 0.9) ✓
▶ tier-4-escape-hatch → 8/8 cases ✓
passed 4/4, failed 0, skipped 0
Closes the 7-PR stack: scaffold, eval harness, 4 tiers of corpus,
references, and now the SKILL.md body that ties everything together.
Adds 11 progressively-disclosed reference files that the skill loads on
demand during translation. Total ~1500 LOC, every file under 200 lines
(skill-creator's progressive-disclosure budget).
api-map.md the comprehensive Remotion -> HF translation table
(the index; loaded at start of translation)
timing.md interpolate, spring (validated configs), easing,
count-up, stagger
sequencing.md Sequence, Series, Loop, Freeze, AbsoluteFill,
Composition root
media.md Audio, Video, Img, IFrame, OffthreadVideo,
staticFile, asset paths
transitions.md @remotion/transitions presentations -> manual GSAP
crossfades or HF shader-transitions
lottie.md @remotion/lottie -> HF lottie adapter (incl. AE
feature limitations note)
fonts.md Google Fonts loading, local @font-face, system
fallback noise floor
parameters.md Zod schemas, defaultProps, sync vs async
calculateMetadata
escape-hatch.md when to bow out + the runtime interop pattern
from PR #214
limitations.md known caveat patterns (volume ramps, Loop with
state, custom presentations, code-split components)
eval.md how to run the validation harness, threshold rule
of thumb, what the noise floor looks like
The references are evidence-driven rather than speculative: every spring
config, easing curve, and SSIM threshold is documented from the
validated T1/T2/T3 calibration runs (mean 0.974 / 0.985 / 0.953). The
escape-hatch boundaries match the lint blockers in PR 2 and the T4
fixtures in PR 5.
Replaces the placeholder .gitkeep from PR 1.
Adds the escape-hatch tier — lint-only fixtures that test the skill's
ability to refuse translation cleanly when it sees patterns that don't map
to HF's seek-driven model.
Cases (8 total):
01-use-state.tsx blocker: r2hf/use-state
02-use-effect-deps.tsx blocker: r2hf/use-effect-deps (multi-line body
with internal commas — regression target for
the regex bug fix in PR 2)
03-async-metadata.tsx blocker: r2hf/async-metadata
04-third-party-react.tsx blocker: r2hf/third-party-react-ui (@mui/material)
05-lambda-config.tsx blocker: r2hf/lambda-import
06-warnings-only.tsx warnings: delayRender / useCallback / useMemo
(no blockers — translates after dropping wrappers)
07-custom-hook.tsx warning: r2hf/custom-hook (pure useFadeIn)
08-mixed.tsx multiple blockers + warnings (aggregate test)
Each case documents:
- The Remotion pattern it demonstrates
- Why it's a blocker / warning / info
- What the skill should do (refuse / drop-and-translate / translate-as-is)
Validation harness (validate.sh):
Runs lint_source.py against each case, asserts:
- Each expected blocker rule fires with severity="blocker"
- Each expected warning rule fires with severity="warning"
- lint_source.py exit code is 1 when blockers expected, 0 otherwise
T4 has no renders to diff. The skill is graded on lint correctness — that's
the gate that decides whether to translate or recommend the runtime interop
pattern from PR #214.
Result: 8/8 cases pass.
Adds the data-driven tier — a purpose-built fixture (option 2 from the
stack discussion, not a port of PR #214's examples/remotion-full/) that
exercises the realistic shape of a production Remotion composition
without using the runtime adapter.
Stargazed.tsx (10s @ 30fps, 1280x720):
Sequence 0-3s TitleScene (title + subtitle)
Sequence 3-7s StatsScene (3 reused StatCards staggered 12 frames apart)
Sequence 7-10s OutroScene (UnderlinedText with scaleX-from-left underline)
Composition shape exercises:
- <Composition schema={z.object({...})} defaultProps={...} />
- nested array prop (stats[]) materialized as repeated HTML
- custom React subcomponents (StatCard, AnimatedNumber, UnderlinedText)
reused with different props
- per-instance delay via prop (delayInFrames -> GSAP timeline offset)
- frame-driven count-up (AnimatedNumber, manual cubic ease-out)
- two different spring configs in the same composition
(damping:12 -> back.out(1.4), damping:14 -> back.out(1.2))
- useCurrentFrame, useVideoConfig
Translation choices documented in README.md and expected.json:
- Zod props -> data-* on root #stage div
- Custom subcomponents inline as repeated HTML using prop interface
as the template
- AnimatedNumber's frame-driven count-up -> GSAP onUpdate tween on a
{ v: 0 } counter object, ease power3.out
- Two different spring configs -> two different back.out overshoots
(1.4 vs 1.2 approximates the damping difference)
- delayInFrames={i * 12} -> GSAP offset (i * 0.4)s
Validated end-to-end: rendered Remotion baseline + HF translation, ran
scripts/render_diff.sh.
measured mean SSIM 0.953
measured min SSIM 0.927
measured p05 SSIM 0.938
threshold 0.90 (~0.04 below p05)
The wider gap vs T1/T2 reflects T3's bigger approximation budget
(2 spring instances + count-up timing + font fallback on multiple text
sizes). Mean SSIM below 0.90 = structural mismatch (wrong durations,
wrong stagger, missing prop wiring), not approximation drift.
Same Remotion config as PR 3: setVideoImageFormat("png") +
setColorSpace("bt709") to match HF's yuv420p output.
Lint: 9 files scanned, 0 blockers / 0 warnings / 0 infos.
oxlint, oxfmt, typecheck all pass.
The fixture is not yet wired into CI; render + diff is documented in
README.md and runs by hand via the harness from PR 2. PR 7's orchestrator
will wire all four tiers into a CI eval run.
Adds the first two test fixtures the skill is graded against. Each fixture
ships:
- remotion-src/ full Remotion project (package.json, src/, remotion.config.ts, tsconfig.json)
- hf-src/ hand-translated HyperFrames composition (index.html)
- expected.json tier metadata + SSIM threshold + translation notes + measured validation
- README.md human walk-through of the translation choices
- setup.sh (T2 only) generates binary assets (PNG, WAV) via ffmpeg
T1 — title-card-fade
- 3 s @ 30 fps, 1280x720
- Single AbsoluteFill, single useCurrentFrame interpolate
with multi-segment input [0,15,75,90] -> [0,1,1,0]
- Validated mean SSIM 0.974, threshold 0.95
(~0.025 gap from font-fallback divergence between Remotion's bundled
Chromium and HF's chrome-headless-shell)
T2 — title-image-outro
- 6 s @ 30 fps, 1280x720, three Sequences (TitleScene, ImageScene, OutroScene)
- Exercises spring, interpolate, Audio, Img, staticFile
- Spring -> GSAP back.out(1.4) translation
- Validated mean SSIM 0.985, threshold 0.95
(translation came out cleaner than predicted; spring->back.out drift was
smaller than the ~0.05 budget I'd expected)
- setup.sh generates a 200x200 blue PNG and a 6 s silent WAV via ffmpeg
so binaries stay out of the repo
Calibration done end-to-end: rendered Remotion baseline + HF translation,
ran scripts/render_diff.sh, set thresholds ~0.02 below measured p05.
Critical Remotion config: setVideoImageFormat("png") + setColorSpace("bt709").
The default JPEG output writes yuvj420p (full-range) which costs ~0.05 SSIM
vs HF's yuv420p (limited-range). Both fixtures' remotion.config.ts encode
this so render_diff.sh measures translation fidelity, not encoder differences.
Both fixtures lint clean (0 blockers via scripts/lint_source.py).
T2 staticFile() references correctly flagged as info-level findings.
The fixtures are not yet wired into CI — that comes with PR 7's orchestrator.
For now, render and eval are documented in each README and run by hand.
Adds the deterministic eval primitives the skill calls into:
scripts/render_diff.sh SSIM diff between two MP4s, JSON summary, configurable threshold
scripts/frame_strip.sh side-by-side comparison strip for visual debugging
scripts/lint_source.py pre-translation lint over Remotion source — blocks/warnings/infos
The harness is decoupled from the render pipeline: it accepts paths to
already-rendered MP4s. The skill orchestrator (PR 7) drives both renders
and feeds the outputs in. This keeps the harness usable in CI, in
sandboxes, and on any machine that has ffmpeg without needing the full
Remotion + HyperFrames toolchain.
Lint catches the patterns from the skill's out-of-scope list:
- useState / useReducer (state-machine driven animation)
- useEffect with deps (side effects)
- async calculateMetadata (Promise-returning composition metadata)
- @remotion/lambda imports
- third-party React UI libraries (MUI, Chakra, Mantine, antd, shadcn, Radix, NextUI)
- delayRender / useCallback / useMemo (warnings)
- staticFile / interpolateColors (info — translatable but flagged)
Smoke test (scripts/tests/smoke.sh) exercises all three scripts against
synthetic inputs: identical ffmpeg testsrc videos pass at threshold 0.99,
different ffmpeg testsrc videos fail at 0.99, frame_strip produces a
strip.png, lint produces 0 blockers on a clean fixture and >=3 blockers
on a fixture that uses useState + useEffect + MUI + async metadata.
Validated locally: smoke.sh exits 0.
Adds the directory + SKILL.md frontmatter for a new skill that translates
Remotion (React) compositions to HyperFrames (HTML+GSAP). This is the
foundation PR; subsequent PRs in the stack add the eval harness, test
corpus, translation references, and finally the SKILL.md body.
The frontmatter description enumerates trigger phrases and explicit
out-of-scope cases (useState/useEffect, async metadata, @remotion/lambda)
so the skill bows out cleanly when a Remotion composition isn't a clean
translation target — those should use the runtime interop pattern from
PR #214 instead.
Validated with skill-creator's package_skill.py.
## What
Forwards the new per-phase extraction breakdown and `tmpPeakBytes` fields from `RenderPerfSummary` (added in #444 and #446) to PostHog via the CLI's existing `render_complete` telemetry event.
## Why
The CLI already ships `render_complete` events to PostHog (`packages/cli/src/telemetry/client.ts`), but `events.ts:trackRenderComplete` only carried a subset of `RenderPerfSummary` — top-level timings, composition dims, and memory snapshots. After #444 added per-phase extraction breakdown (`videoExtractBreakdown`) and #446 added cache hit/miss counters, the data lives on `job.perfSummary` at render-complete but never reaches PostHog dashboards.
Without this, any PostHog insight built around "how often are we hitting the cache?", "what's the median HDR preflight cost?", or "where in the extract phase do compositions spend time?" has to be answered by Datadog log scraping instead.
## How
- **`packages/cli/src/telemetry/events.ts`** — extend `trackRenderComplete` props with 17 new optional fields: `tmpPeakBytes`, the six named stage timings, and the ten `videoExtractBreakdown` fields. All sent as flat properties (`extract_cache_hits`, `stage_capture_ms`, etc.) — PostHog insights query flat keys more ergonomically than nested objects.
- **`packages/cli/src/commands/render.ts`** — wire `job.perfSummary.videoExtractBreakdown` / `stages` / `tmpPeakBytes` into the `trackRenderMetrics` → `trackRenderComplete` hand-off.
- Naming: `extract_phase3_ms` deliberately disambiguates from `stage_video_extract_ms` — the former is just the parallel ffmpeg extract inside Phase 3; the latter is the full stage (resolve + probe + preflight + extract).
- All new fields are optional. The Docker-subprocess branch of `render.ts` that doesn't have a local `perfSummary` still compiles and ships events without them.
## Test plan
- [x] `bun run --cwd packages/cli test` — 161/161 pass
- [x] `bunx tsc -p packages/cli/tsconfig.json --noEmit` — no errors
- [x] `bunx oxlint` + `bunx oxfmt` — clean
- [ ] Once merged, verify PostHog receives the new properties on a real render event (run `hyperframes render` against a fixture and watch PostHog ingestion — telemetry auto-disables in CI, so this requires a local dev render with `HYPERFRAMES_NO_TELEMETRY` unset).
## Stack
Depends on #444 (adds the `videoExtractBreakdown` + `tmpPeakBytes` fields to `RenderPerfSummary`) and transitively on #445 → #446.
## Future work (not in this PR)
- The HeyGen internal producer server (`hyperframes-internal/packages/producer/src/server.ts`) logs `perfSummary` to Datadog via `log.info` but has no PostHog integration. Production renders are the bulk of the traffic — separate PR to either ship perfSummary to PostHog from the internal server, or materialize Datadog log-based metrics for per-phase timings.
## What
Adds a content-addressed cache for extracted video frames, keyed on the tuple `(path, mtime, size, mediaStart, duration, fps, format)`. Repeat renders of the same composition (studio edit → re-render, preview → final) skip the ffmpeg extraction entirely.
## Why
Video frame extraction is the dominant non-capture phase for video-heavy compositions. Studio iteration workflows extract the same frames over and over — each render burns ffmpeg time that adds no value.
Validated on `/tmp/hf-fixtures/cfr-sdr-cache`:
```
Cold (miss): extractMs=69, videoExtractMs=70, totalElapsedMs=2052
Warm (hit): extractMs=1, videoExtractMs=2, totalElapsedMs=1964
cacheHits: 0→1, cacheMisses: 1→0
```
The fixture is tiny (3s CFR SDR @ 30fps), so the wall-clock delta is small; the extraction-time delta (69→1ms, 98%) scales linearly with source length. For heavy-iteration workflows (a user rendering the same composition while tuning encoding params), extraction time goes to zero on every repeat render.
Depends on #444 (instrumentation surface) and #445 (segment-scope HDR preflight — otherwise cache keys would be unstable across renders on mixed-HDR compositions).
## How
- New `packages/engine/src/services/extractionCache.ts`:
- SHA-256 key over a stable JSON encoding of `(path, mtime_ms, size, mediaStart, duration, fps, format)`. Infinity duration is normalized to `-1` so unresolved natural-duration sources still produce stable keys.
- Truncates to 16 hex chars in the entry directory name — 64 bits of entropy is plenty at cache scale and keeps `ls` output short.
- `hfcache-v2-` schema prefix — bumping it invalidates old entries (callers own gc policy; the cache owns keys).
- `.hf-complete` dotfile sentinel. An entry dir without the sentinel is treated as a miss (covers crash-mid-extract and abandoned writes); the next render re-extracts over the partial frames with `-y`.
- `FRAME_FILENAME_PREFIX = "frame_"` shared with the extractor — future refactors only need to touch one place to rename frames.
- `EngineConfig.extractCacheDir` (env: `HYPERFRAMES_EXTRACT_CACHE_DIR`) gates the feature. Undefined disables caching — extraction runs into the render's workDir and cleanup removes it on render end, preserving the prior behaviour exactly. No default root is chosen by the engine; the caller (CLI, app, studio) owns the location policy.
- `ExtractedFrames.ownedByLookup` flag prevents `FrameLookupTable.cleanup` from rm'ing a shared cache dir at render end. Set to `true` on both hits and misses (misses own the directory they wrote into, but hand it over to the cache rather than deleting it).
- Phase 3 extractor flow:
1. Snapshot `(videoPath, mediaStart, start, end)` per resolved video BEFORE Phase 2a/2b preflight mutates them — so cache keys are stable across renders that use workDir-local normalized files (those files have fresh mtimes every render).
2. Compute key, `lookupCacheEntry`.
3. On hit: rebuild `ExtractedFrames` from the cache dir plus the Phase 2-probed `VideoMetadata` — no re-ffprobe.
4. On miss: `ensureCacheEntryDir`, extract with `extractVideoFramesRange(..., outputDirOverride)`, then `markCacheEntryComplete` (the sentinel write is the last step so a crash leaves the dir un-sentineled).
- `extractVideoFramesRange` gains an `outputDirOverride` parameter so cache-miss writes land directly in the keyed dir (no `join(outputDir, videoId)` wrapping).
## Test plan
- [x] 19 unit tests in `extractionCache.test.ts` covering key determinism, mtime/size invalidation, format/fps/mediaStart/duration invalidation, Infinity normalization, sentinel semantics, missing-file tolerance
- [x] 2 integration tests in `videoFrameExtractor.test.ts`:
- "reuses extracted frames on a warm cache hit" — asserts `cacheHits=1`, `extractMs<50ms` on second call against a CFR SDR fixture
- "invalidates the cache when fps changes" — different fps on second call forces a new miss
- [x] End-to-end validation with `HYPERFRAMES_EXTRACT_CACHE_DIR` set, two runs of the same fixture
- [x] Lint + format (oxlint + oxfmt)
- [x] Typecheck (engine + producer)
## What
Scopes the SDR→HDR preflight re-encode to the segment the composition actually uses, mirroring the existing VFR→CFR segment-scope fix.
## Why
`convertSdrToHdr` was re-encoding entire source files, so a 30-minute SDR screen recording contributing a 2-second clip in a mixed HDR/SDR composition ate multi-second preflight time that produced frames no one would ever read. Validated on a mixed 30s-SDR + 2s-HDR fixture: `hdrPreflightMs` drops **87%** (1162→148ms), `videoExtractMs` drops **82%** (1272→231ms), `tmpPeakBytes` drops **45%** (8.2MB→4.5MB).
Depends on #444 (phase-level instrumentation) for the measurement surface.
## How
- `convertSdrToHdr` gains `startTime` and `duration` parameters ahead of the upstream `targetTransfer` arg added by #370. New signature: `convertSdrToHdr(input, output, startTime, duration, targetTransfer, signal, config)`. `-ss $start -t $duration` is added to the ffmpeg args.
- Phase 2 now captures the full `VideoMetadata` per `resolvedVideos` entry (previously just `colorSpace`) so the caller can compute `segDuration` from `video.end - video.start` with a fallback to `metadata.durationSeconds - video.mediaStart` for unbounded (Infinity) clips — without firing another ffprobe.
- After a successful convert, `entry.video.mediaStart` is zeroed out via shallow-copy (doesn't mutate the caller's `VideoElement`) so downstream extraction seeks from 0 instead of the original offset. Mirrors what the VFR→CFR path already does.
## Test plan
Validation on `/tmp/hf-fixtures/hdr-sdr-mixed-scope`:
```
hdrPreflightMs: >1000 → 150 (gate: <300) ✓
videoExtractMs: 1272 → 237 (-82%)
tmpPeakBytes: 8.2MB → 4.5MB (-45%)
```
- [x] Unit test: new regression test synthesizes 10s SDR + 2s HDR fixture inline and asserts the converted file's duration matches the 2s used segment (pre-fix matched the 10s source)
- [x] Lint + format
- [x] Typecheck
- [x] Manual perf validation against synthesized fixture
## What
Adds per-phase timings and counters to `extractAllVideoFrames` and surfaces them on the producer's `RenderPerfSummary` as `videoExtractBreakdown` alongside a new `tmpPeakBytes` workDir size sample.
## Why
Phase 2 video extraction has five distinct sub-phases (resolve, HDR probe, HDR preflight, VFR probe, VFR preflight, per-video extract) and today they collapse into a single `videoExtractMs` stage timing. That makes every subsequent perf PR in this stack immeasurable — you can't tell whether a win came from cache hits, preflight scope reduction, or pure extraction speed.
This PR is foundational for PR #445 (segment-scope HDR preflight) and PR #446 (content-addressed extraction cache).
## How
- New `ExtractionPhaseBreakdown` type with `resolveMs`, `hdrProbeMs`, `hdrPreflightMs/Count`, `vfrProbeMs`, `vfrPreflightMs/Count`, `extractMs`, `cacheHits`, `cacheMisses`. Populated inline with `Date.now()` wrappers — overhead is sub-millisecond on every phase.
- Returned on `ExtractionResult.phaseBreakdown`.
- Producer extends `RenderPerfSummary` with `videoExtractBreakdown?: ExtractionPhaseBreakdown` and `tmpPeakBytes?: number`. `tmpPeakBytes` is sampled from the workDir right before cleanup via a new recursive-size helper that swallows errors (purely observational — a missing workDir must never fail the render).
No changes to the capture-lifecycle resource tracking — earlier versions of this instrumentation plumbed injector LRU stats through `RenderOrchestrator`, which conflicted hard with upstream #371 (`buildHdrCaptureOptions` refactor). Dropped that piece for a marginal observability loss.
## Test plan
Validation on `packages/producer/tests/vfr-screen-recording`:
```json
"videoExtractBreakdown": {
"resolveMs": 0, "hdrProbeMs": 0, "hdrPreflightMs": 0, "hdrPreflightCount": 0,
"vfrProbeMs": 0, "vfrPreflightMs": 166, "vfrPreflightCount": 1,
"extractMs": 97, "cacheHits": 0, "cacheMisses": 0
},
"tmpPeakBytes": 4578598
```
Total elapsed within noise of pre-PR baseline (2665 → 2673 → 3228ms across hosts).
- [x] Unit test: phase-breakdown assertion added to `videoFrameExtractor.test.ts`
- [x] Lint + format (oxlint + oxfmt)
- [x] Typecheck (engine + producer)
- [x] Manual perf validation against VFR fixture
Adds .cursor-plugin/plugin.json at the repo root alongside the existing
.codex-plugin/, so this repo is the single source of truth for the Codex
plugin AND the new Cursor Marketplace submission. Refreshes the shared
assets/logo.png + assets/icon.png to a 1024x1024 / 512x512 symbol-only
mark rasterized from docs/favicon.svg (white background), and renames the
marketplace display title from "HyperFrames" to "HyperFrames by HeyGen"
in both Codex and Cursor manifests.
No skill content changes; purely marketplace-visible branding and the new
Cursor manifest file.
* ci(regression): build test Docker image once, share across shards
Splits regression.yml into a `build-image` job + the existing
`regression-shards` matrix. The build job produces a Docker tarball via
`docker/build-push-action` with `outputs: type=docker,dest=...`, uploads
it as a GHA artifact (retention 1 day, gzip level 1), and each shard
downloads + `docker load`s it instead of rebuilding.
Measured on PR #419 regression runs before the change:
- Docker build step: ~234s per shard WITH GHA layer cache hit
- 11 shards × ~234s = ~43 min of runner time per PR just on redundant
image builds
Cold-cache cases are much worse — happening right now on PR #419 after
release commit b6f50ce bumped every `packages/*/package.json`, invalidating
the COPY layer that feeds `bun install --frozen-lockfile`. All 10 shards
are currently 25-30+ min into a parallel rebuild, thundering-herding
the same npm packages from 10 runners.
After this change:
- 1× build (~4 min warm, ~15 min cold) + 11× (download + `docker load`)
- Expected ~15-20s overhead per shard for artifact download + load
- Net savings: ~30-40 min of runner time per PR run on warm cache,
substantially more on cold cache
The build job doesn't checkout LFS — Dockerfile.test only COPYs source +
package manifests, never the golden baselines, so the image build never
needed LFS. Shards still need LFS for the tests/**/output/output.mp4
baselines they validate against.
* ci(regression): add explicit least-privilege permissions
Addresses CodeQL warning 'Workflow does not contain permissions'.
Defaults the workflow GITHUB_TOKEN to `contents: read` only. The
build-image job elevates to `actions: write` because
`docker/build-push-action` with `cache-from/to: type=gha` uses the
GitHub Actions cache API, which needs read+write on the actions scope.
* docs(readme): note git-lfs requirement for full clones
Repo uses Git LFS for regression-test baselines (~240 MB of .mp4 files
under packages/producer/tests/**/output.mp4). Users cloning without
git-lfs installed hit a cryptic 'git-lfs: command not found' error, as
reported in #407.
Document the requirement with install instructions and the
GIT_LFS_SKIP_SMUDGE=1 escape hatch.
* docs(readme): add Windows install instructions for git-lfs
Per review from @miguel-heygen.
* chore(ci): fix oxfmt formatting on renovate.json
Drive-by to unblock CI. Landed unformatted in #422 because
Renovate's config-migration PR bypasses the lefthook pre-commit hook,
so every subsequent PR's `bun run format:check` (which scans the whole
repo) was failing on this file.
* ci: skip PR runs when targeting a non-main base branch
Adds `branches: [main]` to the `pull_request:` trigger of each workflow
that runs on PRs (CI, regression, Windows render verification, Docs,
Catalog Previews). PRs whose base is something other than main — typical
for stacked PRs — no longer trigger these workflows.
On a 5-PR Graphite stack this turns 5× CI runs into 1× (when the tip
of the stack reaches main). When a child PR is rebased/promoted so its
base becomes main, CI fires as normal.
publish.yml and the default CodeQL setup are untouched: publish already
filters to main, and CodeQL is default-setup (org UI, not a repo YAML).
* chore(ci): fix oxfmt formatting on renovate.json
Same drive-by as #423. Renovate's config-migration PR #422 landed
unformatted (Renovate bot skips lefthook), so every PR branched from
current main fails `bun run format:check`. Whichever of #423 / #426
merges first cleans it up.
Matches the pattern already in place on ci.yml, docs.yml,
windows-render.yml, and catalog-previews.yml. The regression workflow
was the only one without it.
Without this, rapid pushes to a PR leave prior regression runs still
executing their full matrix (~10 parallel shards across styles-a..g,
fast, render-compat, hdr) even though they'll be thrown away. On a busy
day this alone can eat a double-digit share of the GitHub hosted runner
pool and stretch queues for every open PR.
* fix(engine): auto-normalize VFR video inputs to CFR before frame extraction
Screen recordings (macOS ScreenCaptureKit, QuickTime, phone videos) are
commonly variable-frame-rate. When such inputs hit the extractor's
`-ss <start> -i <video> -t <dur> -vf fps=N` pipeline, the fps filter
can emit fewer frames than requested — for a 4-second 30fps segment
starting mid-file, the output was ~90 frames instead of 120.
`FrameLookupTable.getFrameAtTime` returns null for out-of-range indices,
so the compositor held the last valid frame and the user perceived the
video as freezing. This matches the bug report from an X community post
where a user said "all of them freezes" on their screen recording scenes.
The engine already detects VFR via `metadata.isVFR` in ffprobe.ts but
never acted on it — the compiler only logged a warning. This change
mirrors the existing SDR→HDR normalization pattern: when a source is
detected as VFR, re-encode only the used segment with
`-fps_mode cfr -r <fps> -preset fast -crf 18` before extraction.
Scoping the re-encode to `[mediaStart, mediaStart+duration]` means a
30-second clip cut from a 60-minute screen recording pays ~1s of
transcode cost, not 18s. Benchmarked locally:
Baseline (current): 32-39% duplicate frames, 25% frame-count
shortfall on mid-file segments.
Tier 1 (flag changes only): ~same — fps filter issue is not flag-fixable.
Tier 2 (CFR preflight): 1.7-6% duplicate frames, correct frame
count in every scenario tested.
The compiler warning that previously told users to manually re-encode
is downgraded to `console.info` since the engine now handles it.
— Rames Jusso
* refactor(engine): clean up VFR normalization loop after review
- Drop the `vfrNormDirCreated` flag; `mkdirSync({recursive:true})` is
idempotent and cheap.
- Don't re-wrap the `VFR→CFR conversion failed` prefix — `convertVfrToCfr`
already throws a message with that label; adding it again in the catch
produced "VFR→CFR conversion failed: VFR→CFR conversion failed (exit 1)".
- Shorten the Phase 2b header comment; the function docstring above
`convertVfrToCfr` already explains the failure modes and rationale.
- Note which frame windows the VFR fixture's select filter drops so the
magic numbers are scannable.
No behavior change; 311/311 engine tests still pass.
— Rames Jusso
* test(engine): add VFR regression unit tests
Adds a describe block that synthesizes a VFR fixture via ffmpeg and asserts
the extractor produces the expected frame count (no shortfall) and no long
runs of duplicate frames — the user-visible "frozen screen recording"
symptom. Covers both a mid-file segment and the full-file case.
Guarded with describe.skipIf(!HAS_FFMPEG) because the CI Test job on
ubuntu-24.04 and the Windows test-windows job don't install ffmpeg. The
producer-level regression test in packages/producer/tests/vfr-screen-recording/
runs inside Dockerfile.test (which has ffmpeg) and is the primary CI signal
for this bug; these unit tests are supplementary coverage for local and
any ffmpeg-equipped CI environment.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(producer): add vfr-screen-recording regression test
End-to-end CI regression coverage for PR #360 via the existing
regression-harness: renders a 3s composition containing a real macOS
ScreenCaptureKit clip (r_frame_rate=120, avg≈36fps) seeked to
mediaStart=1, then PSNR-compares against a committed output.mp4.
Fixture src/clip.mp4 (108 KB) is a 5-second excerpt downscaled to 480×332
with -fps_mode passthrough to preserve the VFR timestamps. Content is the
public hyperframes OSS repo root page — see NOTICE.md for provenance.
With the fix applied, all 100 PSNR checkpoints pass. With the fix reverted,
66 of 100 fail (PSNR drops from ~43 dB to ~20 dB in the duplicate-frame
windows). Tagged "regression,video,vfr" so it runs in the fast shard
of .github/workflows/regression.yml automatically.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(producer): regenerate vfr-screen-recording baseline in Docker
The committed golden output.mp4 was initially rendered on the host machine;
CI runs the renderer inside Dockerfile.test with a different Chrome +
ffmpeg build, producing pixel-level drift that failed PSNR at 54/100
checkpoints (~20 dB vs 41 dB in the VFR sparse-content windows). Both
renders are valid — the VFR source has inherent sampling ambiguity in
static segments, and different Chrome/ffmpeg builds make different valid
choices.
Regenerated the baseline via `bun run docker:test:update vfr-screen-recording`
so it matches the Docker environment CI actually uses. Matches the flow
the existing sub-composition-video, hdr-pq, etc. baselines were captured
with.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs: document that producer test baselines must be captured in Docker
Hit this 2026-04-21 with the vfr-screen-recording regression test:
host-generated output.mp4 baseline tripped 54/100 PSNR checkpoints in CI
because Chrome + ffmpeg drift between the host and Dockerfile.test.
Document the `bun run --cwd packages/producer docker:test:update <name>`
flow so future contributors don't repeat the mistake.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(player): inject runtime immediately for nested compositions
Compositions that use `data-composition-src` on child elements require
the HyperFrames runtime to load those scenes — there is no way for the
iframe to render without it. The existing probe loop delayed runtime
injection behind a 5-tick attempts gate so the adapter path could try
to resolve a timeline first.
For nested compositions that race lost: a composition like the
`product-promo` registry example registers an inline pre-runtime GSAP
timeline at `window.__timelines["main"]` (covering only a partial
duration, e.g. 14s of a 20s master) while the iframe document loads.
The probe's adapter check finds that timeline and locks the player into
a "ready" state against it — which short-circuits the attempts gate and
the runtime never gets injected. The iframe ends up blank because the
runtime is what would have loaded the child scenes via
`data-composition-src`.
This change splits the injection decision into a pure helper,
`shouldInjectRuntime(state)`, and treats nested compositions as
"inject immediately, skip the gate." Self-contained GSAP-only
compositions retain the 5-tick grace period so the adapter path keeps
first shot for them.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(core): propagate play/pause to all sibling timelines
Pausing or playing the master timeline only called `.pause()` / `.play()`
on `state.capturedTimeline` — the single adapter-selected timeline. In a
nested composition (a master with `data-composition-src` children), each
scene's own timeline is registered as a sibling in `window.__timelines`,
so they would keep advancing after the user clicked pause. The player UI
froze at the paused time while the visual content continued to animate,
eventually finishing all scene-level animations and landing on an empty
end-state.
Wire `window.__timelines` into the runtime player via a new
`getTimelineRegistry` dep, iterate the registry on play/pause, and
forward `timeScale` to siblings when play() starts so a changed
playback-rate applies uniformly.
Covered by 7 new unit tests in player.test.ts, including the identity-
equality check (don't double-invoke the master), playbackRate
propagation, a broken-sibling swallow, and a back-compat case with no
registry supplied.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds honest Hyperframes vs Remotion comparison: README section with
paragraph + table + open-source-vs-source-available callout, plus a
full guide at docs/guides/hyperframes-vs-remotion.mdx walking through
the core React-vs-HTML decision, practical differences (including a
GSAP side-by-side), and licensing.
Closes#318
* feat(cli): add --lang and auto-infer phonemizer locale from voice prefix
`hyperframes tts` was calling Kokoro's `model.create(text, voice=, speed=)`
with no language argument, so Kokoro's default phonemizer (en-us) was
applied regardless of the voice selected. Picking `ef_dora` or `jf_alpha`
and feeding it Spanish or Japanese text produced English-phonemized
output.
Closes#349.
- `manager.ts`: add `SUPPORTED_LANGS`, `inferLangFromVoiceId`, and
`isSupportedLang`. Attach a `defaultLang` field to every bundled voice
and expand the bundled list with `ef_dora`, `ff_siwis`, `jf_alpha`,
`zf_xiaobei` so `--list` surfaces multilingual options.
- `synthesize.ts`: accept optional `lang: SupportedLang` in
`SynthesizeOptions`, forward it to the Python worker as `argv[7]`.
The worker introspects `Kokoro.create`'s signature and only passes
`lang=` when the installed kokoro-onnx version supports it. Returned
metadata now includes `lang` and `langApplied` so callers can detect
silent no-ops. Bump the cached script filename to `synth-v2.py` so
existing installs pick up the new script automatically.
- `commands/tts.ts`: add `--lang, -l` with validation against
`SUPPORTED_LANGS`. Resolution order is explicit `--lang` > inferred
from voice prefix > `en-us`. When explicit lang disagrees with the
voice-implied lang (legitimate for stylized accents), emit a
dim-level hint; suppress under `--json`. When kokoro-onnx silently
ignores the kwarg, log that too. Update `--list` with a new
"Lang code" column and add multilingual examples.
- Tests: new `manager.test.ts` covering every supported prefix, the
unknown-prefix fallback, case-insensitivity, `isSupportedLang`
validation, and a regression guard that every bundled voice has a
valid `defaultLang` matching its ID.
- Docs: `docs/packages/cli.mdx` and `skills/hyperframes/references/tts.md`
updated with the flag, examples, the espeak-ng dependency note for
non-English phonemization, and the voice-prefix → lang table.
Backward compatibility:
- English voices (a*/b* prefixes) continue to phonemize as en-us / en-gb
— no change.
- Non-English voices now phonemize correctly by default (bug fix, not a
regression).
- Older kokoro-onnx versions that don't know the `lang` kwarg keep
working via signature introspection; the CLI logs a dim note if
`--lang` was requested but ignored.
Verification:
- `bun --cwd packages/cli test` — 128 tests pass (incl. 17 new).
- `bunx oxlint` and `bunx oxfmt --check` clean on changed files.
- `bun run build` succeeds.
- `npx tsx packages/cli/src/cli.ts tts --help` / `--list` render cleanly;
invalid `--lang` produces a clean error with the valid-codes list.
* refactor(cli): simplify tts --lang implementation
Post-review cleanup on #351. Net -21 lines.
- Drop `defaultLang` field + `makeVoice()` helper from VoiceInfo —
compute via `inferLangFromVoiceId(v.id)` at read time in listVoices.
The only reader was the --list table; caching the derived value on
every voice added a self-consistency invariant we had to test.
- Drop redundant `lang` field from SynthesizeResult — caller already
knows the requested lang since it passed it in; only `langApplied`
carries information the caller can't derive.
- Use `errorBox` for --lang validation to match the house style in
render.ts (other validation errors already use errorBox).
- Reuse existing `langList` module constant in the validation error
instead of re-joining SUPPORTED_LANGS.
- Inline `DEFAULT_LANG` — used once in inferLangFromVoiceId.
- Trim WHAT-restating comments and the duplicate prefix-enumeration
JSDoc on inferLangFromVoiceId (VOICE_PREFIX_LANG already carries
per-row comments).
- Clean up orphaned `synth*.py` files in ~/.cache/hyperframes/tts
when writing the current versioned script, so repeated upgrades
don't leak files.
- Drop the `EN-US` case-sensitive-rejection test assertion — the CLI
lowercases input before validation, so accepting mixed case is a
feature, not a bug.
Tests: 16/16 in `manager.test.ts`, 127/127 full CLI suite pass.
Lint + format + typecheck clean.
* refactor(engine): restructure frame reorder buffer with Map-keyed storage
Rewrites createFrameReorderBuffer to use a Map<number, Array<() => void>>
keyed by frame index instead of a flat Array<{frame, resolve}> scanned on
every advance. O(1) lookups in enqueue/flush, fast-paths for the matching-
cursor and overshoot cases, and a small fix: waitForAllDone now coexists
with the writer still waiting on the final frame instead of colliding on
the same waiter slot.
Also adds 5 unit tests (there were none before) covering the fast-path,
out-of-order gating, multi-waiter-per-frame semantics, waitForAllDone
normal path, and the overshoot case.
Comment tweaks on buildChromeArgs — the flag profile is the standard
headless-for-capture set (Puppeteer / Playwright / Chrome headless-shell
all converge on similar flags); rephrased for clarity.
* refactor(cli): simplify port availability probe with async/await
Rewrites isPortAvailableOnHost from a single new-Promise callback into an
async/await form with an intermediate `bindError: ErrnoException | null`
variable. Makes the bind-then-release flow explicit as two sequential
awaits, and broadens the non-EADDRINUSE errno commentary (EADDRNOTAVAIL
for disabled IPv6, EACCES for privileged ports, EAFNOSUPPORT for missing
address families — all treated as "this host doesn't apply", not "port
occupied").
No behavior change to existing callers; all four portUtils tests still
pass.
* docs: add CREDITS.md and surface website-to-hyperframes skill
- New CREDITS.md acknowledging prior art in the browser-based video
rendering space (Remotion) and the ecosystem HyperFrames builds on
(Puppeteer, FFmpeg, GSAP, Hono). Standard OSS practice.
- Adds the `website-to-hyperframes` skill to the skills tables in
README.md, docs/guides/prompting.mdx, and the project template at
packages/cli/src/templates/_shared/CLAUDE.md. The skill ships in
skills/ but was missing from every table.
- Adds `/hyperframes-registry` to the prose mention in the repo
CLAUDE.md.
* fix(core): drive adapter seeks when composition has no GSAP timeline
renderSeek returned early when deps.getTimeline() was null, skipping the
onDeterministicSeek call that drives all frame adapters (CSS, WAAPI,
Lottie, Three.js). That meant compositions using any non-GSAP animation
primitive froze on their initial frame during capture.
Now we still quantize the seek time and fire onDeterministicSeek even
without a timeline, so each adapter gets a chance to advance.
GSAP compositions are unaffected — timeline-driven seek still takes the
same path it did before.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(producer): auto-fallback screenshot capture for raf and iframes
Co-Authored-By: Codex <codex@openai.com>
* test(producer): add render compatibility regression fixtures
Co-Authored-By: Codex <codex@openai.com>
* fix(core): scrub CSS animations via WAAPI currentTime
Co-Authored-By: Codex <codex@openai.com>
* test(producer): cover css keyframe renders
Co-Authored-By: Codex <codex@openai.com>
* fix(producer): propagate virtual time into iframe documents
Co-Authored-By: Codex <codex@openai.com>
* test(producer): refresh iframe docker golden
Co-Authored-By: Codex <codex@openai.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Codex <codex@openai.com>
* docs(guides): add performance guide and preview-stutter troubleshooting
Adds a dedicated Performance guide covering preview-vs-render cost model,
expensive CSS patterns (backdrop-filter, filter, shadows), image sizing,
and how to diagnose slow compositions with Chrome DevTools.
Cross-links from troubleshooting (new "Preview stutters" accordion) and
common-mistakes (new "Oversized source images" and "Heavy backdrop-filter
stacks" accordions). Wires the new page into docs.json nav.
Also fixes a pre-commit format hook edge case: oxfmt would exit 2 when
the only staged files matching the format glob were all covered by
.prettierignore (e.g. docs-only changes). Add --no-error-on-unmatched-pattern
to the lefthook oxfmt invocation so docs-only commits are not blocked.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs: call out preview performance limits at the entry points
The preview command, studio package, and determinism concept pages all
frame preview as visually equivalent to render — correct for fidelity,
misleading for playback smoothness. A user who reads those pages and
then hits a paint-heavy composition has no way to know why preview
stutters, short of drilling into troubleshooting.
Adds short notes at each entry point linking out to the new Performance
guide, so users hit the "preview is hardware-bound, render isn't"
explanation wherever they land first.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
HTML files in this repo are compositions (user-facing content and
registry templates), not the framework source. The framework itself is
TypeScript. Hide HTML from Linguist's detection so the language bar on
GitHub reflects what the repo actually implements.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Chrome's "Failed to load resource" message text does not include the failing
URL — it's only on msg.location().url. The previous filter in frameCapture.ts
only checked msg.text(), so every font 404 (e.g. Google Fonts <link> tags
in sandboxed render environments) fell through to the "[non-blocking]"
prefix instead of being suppressed.
Extract the classifier into isFontResourceError() and match against both
text and location.url, and extend the extension match to .ttf/.otf. Adds
a unit test covering the URL-in-location, URL-in-text, and non-font cases.
This is a targeted fix for the render-output noise that PR #311 attempted
to address by adding a ~120-entry SYSTEM_FONTS skip list. That approach
silently shadowed existing FONT_ALIASES (arial→inter, helvetica→inter,
courier new→jetbrains-mono, segoe ui→roboto, etc.) and changed render
output on Linux fleets that don't have those fonts installed. Fixing the
console-log filter here suppresses the noise without changing any font
resolution behavior.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Move all preview mp4/png/gif assets under docs/images/ out of the repo
and serve them from https://static.heygen.ai/hyperframes-oss/docs/images/
(backed by s3://heygen-public/hyperframes-oss/docs/images/, CloudFront).
Drops ~49MB from the working tree and, more importantly, ~49MB from every
future Mintlify build checkout. Combined with the (already-LFS-tracked)
producer snapshots, the remaining bloat in 'npx skills add heygen-com/
hyperframes' (see #300) is LFS smudge during clone — separate fix needed
in the skills CLI to pass GIT_LFS_SKIP_SMUDGE=1.
Changes:
- Delete docs/images/** (103 files, ~49MB). Files are uploaded to S3 already.
- Rewrite /images/* references in 44 MDX files, TemplateCard.jsx, and
catalog-index.json to absolute CDN URLs.
- Update README.md img src to CDN URL (renders correctly on GitHub).
- Add docs/images/ to .gitignore so regenerated previews aren't committed.
- Add scripts/upload-docs-images.sh to sync docs/images/ → S3 after running
the preview generators.
- Wire up bun run upload:docs-images and bun run generate:catalog-previews
scripts in package.json.
- Update generator script docstrings to point at the upload step.
External contributors can still regenerate previews locally (mintlify dev
reads the CDN URLs, so broken previews appear only for newly added items
pending a maintainer upload). Maintainers run:
bun run generate:catalog-previews --only <name>
bun run upload:docs-images
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Raise default encoding quality to visually lossless at 1080p (CRF 18)
and expose fine-grained encoding controls for power users.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: add prompt cookbook + prompting guide for AI agents
Addresses user feedback that there's no guidance on how to actually
prompt Claude Code (or other agents) once the hyperframes skills are
installed. Adds copy-pasteable example prompts in the README and
quickstart, a new prompting guide page, and a starter-prompt nudge in
the `hyperframes init` output.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(prompting): add vocabulary tables, rules, and TTS voice guide
Merges the best content from the internal prompt guide into
prompting.mdx: easing vocabulary, caption tone table, transition
energy matrix, audio-reactive frequency mapping, marker highlight
modes, TTS voice recommendations, rendering quality presets, and
framework rules (technical requirements vs best practices).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(prompting): rename page title to "Prompt Guide"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* chore: remove greensock/gsap-skills dependency, fix Math.random nuance
The bundled skills/gsap/ already covers the GSAP surface needed for
HyperFrames compositions. Installing greensock/gsap-skills on top adds
a competing full-ecosystem skill that's mostly irrelevant (ScrollTrigger,
Draggable, SplitText, etc.) and can confuse agents about which GSAP
context to load.
Also adds seeded-PRNG nuance to the Math.random() rule in the prompt
guide (matching the skill's actual guidance).
Removed from: skills.ts, README, AGENTS.md, shared AGENTS.md/CLAUDE.md,
and prompting.mdx.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* chore: require minimal reproduction link in bug report template
Adds a required "Link to reproduction" input field asking users to push
a minimal repro to a public GitHub repo (scaffolded via
`hyperframes init repro --non-interactive --example blank`).
Also consolidates the OS/Node/FFmpeg/version fields into a single
"Environment" field using `npx hyperframes info` output — fewer fields
to fill, more consistent data.
Follows the same pattern as Next.js and Gatsby issue templates.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(issue-template): use hyperframes doctor for environment info
`hyperframes info` only prints project metadata (resolution, duration,
elements). `hyperframes doctor` prints the full environment: version,
Node.js, FFmpeg, Chrome, memory, disk, Docker — everything needed to
diagnose bugs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(prompting): mention validate alongside lint in anti-patterns
Per Vance's review comment — validate catches runtime errors (JS
exceptions, missing assets, contrast) that lint doesn't.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: replace libretto example URL with hyperframes repo
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The `Copy page` dropdown panel rendered with a transparent background in
light mode because `docs/custom.css` defined `--background-light: #ffffff`
on `:root`. Mintlify's Maple theme owns that variable as a Tailwind color
(space-separated RGB used via `rgb(var(--background-light)/<alpha>)`), so
the hex override produced invalid CSS like `rgb(#ffffff/1)` and the
dropdown's `bg-background-light` class fell back to transparent. Dark
mode was unaffected because the dropdown panel uses `bg-background-dark`,
which custom.css didn't redefine.
Namespaced every custom variable with `--hf-` to make collisions
impossible, and updated the two consumers (`pre`, `::selection`, link
color in custom.css; `.tpl-card:hover` border in template-gallery.css).
## What
Fixes the README skills table to match actual skill names, and moves two orphaned script directories into the `hyperframes` skill where they belong.
## Why
**README**: Listed `hyperframes-compose` and `hyperframes-captions` as separate skills — these don't exist. Captions/compose are part of the `hyperframes` skill. Also listed `gsap-core, gsap-timeline, gsap-plugins, ...` but the actual skill is just `gsap`. Missing `hyperframes-cli` entirely.
**Orphaned scripts**: `skills/hyperframes-animation-map/` and `skills/hyperframes-contrast/` had scripts but no `SKILL.md` — they looked like broken skills and wouldn't be installed by `npx skills add`. They're helper scripts invoked by the main `hyperframes` skill (SKILL.md already references them in the "Quality Checks" section). Moving them under `skills/hyperframes/scripts/` makes them part of the skill they belong to.
## How
**README skills table** — corrected to match the 4 actual skills:
- `hyperframes` (was `hyperframes-compose` + `hyperframes-captions`)
- `hyperframes-cli` (was missing)
- `hyperframes-registry` (unchanged)
- `gsap` (was `gsap-core, gsap-timeline, gsap-plugins, ...`)
**Script moves:**
- `skills/hyperframes-animation-map/scripts/animation-map.mjs` → `skills/hyperframes/scripts/`
- `skills/hyperframes-contrast/scripts/contrast-report.mjs` → `skills/hyperframes/scripts/`
- Removed empty `skills/hyperframes-animation-map/` and `skills/hyperframes-contrast/`
- Updated path references in SKILL.md, both script headers, and `contrast-audit.browser.js`
## Test plan
- [ ] `grep -r "hyperframes-animation-map\|hyperframes-contrast" --include="*.md" --include="*.mjs" --include="*.js" --include="*.ts" .` returns no results
- [ ] `ls skills/` shows only `gsap`, `hyperframes`, `hyperframes-cli`, `hyperframes-registry`
- [ ] README skills table matches `skills/*/SKILL.md` names
- [x] Documentation updated (if applicable)
## What
Restructures the Quickstart docs page to lead with AI agent onboarding as the recommended path, matching the README and homepage flow.
## Why
The README (PR #277) now leads with skills-first onboarding, but the Quickstart docs page still led with `npx hyperframes init`. This creates a consistency gap — someone clicking "Quickstart" from the README would see a different onboarding flow than what they just read.
## How
- **Option 1 (recommended)**: Install skills → prompt your agent → iterate by describing changes
- **Option 2**: Manual CLI setup (`hyperframes init` → preview → edit → render) — unchanged content, now under a sub-heading
- Added tip explaining why skills matter (framework-specific patterns)
- Notes that `hyperframes init` installs skills automatically
- "Next steps" cards now include the Catalog (50+ blocks) replacing the Compositions card
## Test plan
- [ ] Preview the Mintlify docs and verify the Quickstart page renders correctly
- [ ] Verify Option 1 flow reads naturally for someone new to HyperFrames
- [ ] Verify Option 2 manual flow is unchanged (same steps, same code examples)
- [ ] Verify all links resolve (Examples, Catalog, GSAP Animation, Rendering)
- [x] Documentation updated (if applicable)
## What
Adds `AGENTS.md` at two levels:
1. **Repo-level** (`/AGENTS.md`) — for contributors working on HyperFrames itself
2. **Project-level** (`packages/cli/src/templates/_shared/AGENTS.md`) — scaffolded into user projects by `hyperframes init`
Also replaces the previous `AGENTS.md → CLAUDE.md` symlink with a standalone file.
## Why
AGENTS.md is the emerging universal standard for AI coding tool configuration, supported by Claude Code, Cursor, GitHub Copilot, Gemini CLI, and Codex (60K+ repos). Remotion already has one. For a project that positions itself as AI-native, this is a gap.
The project-level file (scaffolded by `init`) ensures every AI tool — not just Claude — gets framework context when working on a user's composition project. The symlink was replaced because symlinks are fragile on Windows and the content should differ (repo-level covers build/test, project-level covers composition rules).
## How
- **Repo-level AGENTS.md**: Build/test/lint commands, project structure, key conventions, skills install, doc links
- **Project-level AGENTS.md**: Composition-specific — skills, CLI commands, project structure, linting workflow, key rules, doc links
- **CLAUDE.md** remains for Claude-specific skill invocation syntax (slash commands)
## Test plan
- [ ] Verify AGENTS.md renders correctly on GitHub
- [ ] Verify no duplication with CLAUDE.md (AGENTS.md = universal basics, CLAUDE.md = Claude-specific slash commands)
- [ ] Manual testing performed
- [x] Documentation updated (if applicable)
## What
Restructures the README to lead with skills-first onboarding, adds a demo GIF, surfaces the catalog, fixes incorrect pnpm references in contributing docs, and corrects the HTML example to use actual attribute names.
## Why
The README told a CLI-first story while the homepage (hyperframes.heygen.com) tells an AI-agent-first story. For a project that brands itself "built for agents," the GitHub landing page should match. Additionally, 50+ catalog blocks were invisible from GitHub, the player and shader-transitions packages were missing from the packages table, and the contributing docs referenced pnpm while the repo uses bun.
## How
**README changes:**
- Quick Start restructured: skills install as Option 1 (recommended), manual CLI as Option 2
- Added demo GIF rendered with HyperFrames itself (HTML + GSAP composition → MP4 → GIF)
- Added Catalog section with install examples and link
- Added `@hyperframes/player` and `@hyperframes/shader-transitions` to packages table
- Added npm downloads badge
- Fixed HTML example: `data-track` → `data-track-index`, added missing `class="clip"` on img
- Condensed Skills section into a concise table
- Documentation link now points to `/introduction` (Mintlify docs) instead of the landing page
**testing-local-changes.mdx:**
- All `pnpm` references replaced with `bun` (14 occurrences)
- Path references updated from `hyperframes-oss` to `hyperframes`
## Test plan
- [ ] Verify README renders correctly on GitHub (logo, badges, GIF, tables, code blocks)
- [ ] Verify GIF loops and is readable at GitHub's default README width
- [ ] Verify all links resolve (docs site, catalog, packages, contributing)
- [ ] Read through testing-local-changes.mdx for any remaining pnpm references
- [x] Documentation updated (if applicable)
Replace old text-only wordmarks and complex favicon with new HyperFrames
brand assets featuring the gradient icon. Add dark/light logo switching
to README via GitHub's <picture> element.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds `hyperframes catalog` for browsing the registry:
- Default: non-interactive table output (agent-friendly)
- --type block/component and --tag filters
- --json for machine-readable output
- --human-friendly for interactive picker that installs on select
Registered in cli.ts, help.ts, documented in docs/packages/cli.mdx.
## What
Add 28 transition blocks from the Hyperframe Template Structure catalog, bringing the registry to 53 total items.
### Shader transitions (14 blocks, WebGL, 4s each)
`domain-warp-dissolve`, `ridged-burn`, `whip-pan`, `sdf-iris`, `ripple-waves`, `gravitational-lens`, `cinematic-zoom`, `chromatic-radial-split`, `glitch`, `swirl-vortex`, `thermal-distortion`, `flash-through-white`, `cross-warp-morph`, `light-leak`
### CSS transition showcases (14 blocks, various durations)
`transitions-3d`, `transitions-blur`, `transitions-cover`, `transitions-destruction`, `transitions-dissolve`, `transitions-distortion`, `transitions-grid`, `transitions-light`, `transitions-mechanical`, `transitions-other`, `transitions-push`, `transitions-radial`, `transitions-scale`, `transitions-shader`
## Why
Phase D content accumulation. Transitions are the most-requested category for the catalog.
## How
- Shader transitions extracted from `shader-showcase.zip`, each a standalone HTML with WebGL shaders
- CSS transitions extracted from `showcase-bundle.zip`, each a standalone showcase page
- All tagged with `transition` + `shader` or `showcase` for catalog grouping
- Preview thumbnails generated for all 28 blocks
- Catalog pages + index regenerated
## Test plan
- [x] All 28 blocks produce preview thumbnails
- [x] `registry-item.json` validates for all blocks
- [x] Catalog pages generated (45 total items in catalog-index.json)
- [x] `oxfmt --check` passes
## What
New skill `hyperframes-registry` that teaches AI coding agents how to install and wire registry blocks and components into HyperFrames compositions.
### Skill structure
```
skills/hyperframes-registry/
SKILL.md — triggers, overview, quick reference
references/
install-locations.md — default paths, hyperframes.json config
wiring-blocks.md — iframe inclusion, data attributes, positioning
wiring-components.md — snippet merging (HTML, CSS, JS, timeline)
discovery.md — manifest reading, item fields, available items table
demo-html-pattern.md — why components ship demo.html, structure conventions
examples/
add-block.md — worked example: data-chart block install + wiring
add-component.md — worked example: shimmer-sweep component install + wiring
```
## Why
Phase B of the catalog plan (PR 10). Without this skill, agents using `hyperframes add` have to guess how to wire installed items into compositions. The skill encodes the iframe/snippet patterns so agents get it right on the first attempt.
## How
- SKILL.md frontmatter triggers on: `hyperframes add`, "block", "component", `hyperframes.json`
- References cover every step: discovery, install, wiring blocks (iframe), wiring components (snippet merge), and the demo.html convention
- Two worked examples walk through complete install-to-preview workflows
- Updated CLAUDE.md skills table + trigger rules, README.md skills table, docs/packages/cli.mdx
## Test plan
- [x] `scripts/lint-skills.ts` passes (checked 4 skill files, no issues)
- [x] `oxfmt --check` passes on all markdown files
- [x] SKILL.md frontmatter has valid `name` and `description`
- [x] All reference links in SKILL.md resolve to existing files
- [x] CLAUDE.md, README.md, and docs CLI page updated with new skill
## What
Script that auto-generates per-item catalog documentation from `registry-item.json` manifests.
**New file:** `scripts/generate-catalog-pages.ts`
**Outputs:**
- `docs/catalog/blocks/<name>.mdx` — per-block detail page
- `docs/catalog/components/<name>.mdx` — per-component detail page
- `docs/public/catalog-index.json` — flat manifest for the grid page (constant-sized regardless of catalog size)
- Updates `docs/docs.json` with a Catalog tab containing Blocks + Components groups
## Why
Phase B of the catalog plan (PR 9). After this lands, future content PRs don't need to write MDX by hand — the script generates everything from `registry-item.json`.
## How
The script:
1. Walks `registry/blocks/*/registry-item.json` and `registry/components/*/registry-item.json`
2. **Wipes `docs/catalog/` before regenerating** — deleted items don't leave stale pages
3. Generates MDX per item with: title, description, tag badges, preview image, install command, details table, files table, usage hint, and related skill link
4. Emits `catalog-index.json` with `{name, type, title, description, tags, href, preview}` per item
5. Updates `docs.json` navigation — inserts or replaces the Catalog tab with current block/component page lists
Run before Mintlify builds: `npx tsx scripts/generate-catalog-pages.ts`
## Test plan
- [x] Script compiles — passes `lefthook` typecheck + lint + format
- [x] CONTRIBUTING.md documents the auto-generation workflow
- [ ] Full end-to-end test requires PRs 6+7 to merge first (items must exist in registry/)
## What
CI workflow that auto-renders preview thumbnails for new/changed registry blocks and components on pull requests.
**New files:**
- `scripts/generate-catalog-previews.ts` — catalog preview renderer supporting all three registry item types
- `.github/workflows/catalog-previews.yml` — GitHub Actions workflow triggered on PRs touching `registry/blocks/` or `registry/components/`
## Why
Phase B of the catalog plan (PR 8). After this lands, future block/component PRs don't need to manually generate preview images — CI handles it automatically.
## How
The preview script discovers items from the registry directory structure:
- **Examples**: renders `index.html` (same as the existing `generate-template-previews.ts`)
- **Blocks**: renders the block's standalone HTML file directly (e.g., `data-chart.html`)
- **Components**: renders the component's `demo.html` (the demo.html convention from PR 7)
The CI workflow:
1. Detects which blocks/components changed in the PR via `git diff`
2. Renders thumbnails for only the changed items (not the full catalog)
3. Uploads preview PNGs as artifacts
Output goes to `docs/images/catalog/<type>/<name>.{png,mp4}` (separate from the existing `docs/images/templates/` directory).
Supports CLI flags: `--only <name>`, `--type <example|block|component>`, `--skip-video`.
## Test plan
- [x] Script compiles and passes typecheck (`lefthook pre-commit` ran lint + typecheck + format)
- [x] Workflow YAML is valid (standard GitHub Actions syntax, follows existing ci.yml patterns)
- [ ] Full end-to-end test requires Chrome + FFmpeg (runs in CI, not testable locally without producer deps)
## What
Three reusable effect components for the registry, each with a snippet HTML and companion `demo.html`:
| Component | Description |
|-----------|-------------|
| `grain-overlay` | Animated film grain texture overlay (CSS keyframes, extracted from warm-grain example) |
| `shimmer-sweep` | CSS gradient light sweep across text/elements, driven by GSAP custom property animation |
| `grid-pixelate-wipe` | Grid-based dissolve transition — screen breaks into 16×9 squares that scale in/out with stagger |
Establishes the `demo.html` convention in `CONTRIBUTING.md`.
## Why
Phase B of the catalog plan — seed the first components in the registry. Components are effect snippets that get merged into existing compositions (vs. blocks which are standalone sub-compositions).
## How
- **grain-overlay**: Extracted the grain texture pattern from the warm-grain example. Uses a 200% oversized tiled texture with `steps(1)` keyframe animation for the random-noise effect.
- **shimmer-sweep**: Original implementation using CSS custom properties (`--shimmer-pos`) animated by GSAP. The gradient mask uses `mix-blend-mode: overlay` for a natural light sweep. Auto-injects `.shimmer-mask` elements into `.shimmer-sweep-target` wrappers.
- **grid-pixelate-wipe**: Creates a 16×9 CSS Grid of cells, animated with GSAP stagger. Users drive `.grid-cell` `scale` directly in their timeline.
Simplify review addressed: scoped `.grain-texture` under `#grain-overlay`, scoped `.grid-cell` under `#grid-pixelate-overlay`, removed `window.gridPixelateIn/Out` globals in favor of direct GSAP patterns.
Each component ships a `demo.html` — a standalone composition that previews the effect and doubles as a fixture for the CI preview pipeline (PR 8).
## Test plan
- [x] `hyperframes add grain-overlay` installs to `compositions/components/grain-overlay.html`
- [x] `hyperframes add shimmer-sweep` installs to `compositions/components/shimmer-sweep.html`
- [x] `hyperframes add grid-pixelate-wipe` installs to `compositions/components/grid-pixelate-wipe.html`
- [x] All three return correct `--json` output with snippet and type info
- [x] `registry-item.json` files validate against the JSON Schema
- [x] `demo.html` files are self-contained with correct `data-composition-id` and `window.__timelines` registration
- [x] `oxfmt --check` and `oxlint` pass on all files
- [x] `CONTRIBUTING.md` documents the `demo.html` convention and registry item checklist
Extract three standalone blocks from existing example templates:
- data-chart: Animated bar + line chart with staggered reveal (from nyt-graph)
- flowchart: Decision tree with SVG connectors and cursor interaction (from decision-tree)
- logo-outro: Cinematic logo reveal with tagline and URL pill (from product-promo)
Each block is a complete HTML composition installable via `hyperframes add <name>`.
Blocks render as iframes in host compositions.
* fix(player): parent-frame media playback for mobile
Mobile browsers block media.play() inside iframes when the user
gesture happened in the parent frame — postMessage doesn't transfer
user activation (per the User Activation v2 spec).
## Problem
The player renders compositions in a sandboxed iframe. When a user
taps play in the parent frame, the player sends a postMessage to the
iframe's runtime, which calls audio.play(). On mobile, this fails
silently because the iframe has no user activation context.
## Solution
The player now extracts ALL timed media elements (audio/video with
data-start) from the iframe's DOM (same-origin access), creates
parent-frame copies, and disables the iframe originals. On play(),
parentMedia.play() runs synchronously in the gesture call stack,
satisfying mobile autoplay policy.
### Generic media handling
- Finds all `audio[data-start], video[data-start]` in the iframe
- Creates a parent-frame copy for each (Audio or Video element)
- Preserves data-start offsets for correct seek positioning
- Strips data-start from iframe elements so the runtime ignores them
- Falls back to iframe media for cross-origin iframes
### `audio-src` attribute
Convenience for the common single-narration case. When set, the
player starts preloading audio immediately — before the iframe loads.
This eliminates the loading delay that caused jittery playback.
### No active sync
Both parent media and the GSAP timeline are real-time systems. When
started simultaneously, they naturally stay within ~10ms — no drift
correction needed. Active sync with coarse granularity (50ms polling)
caused MORE jitter than it prevented via repeated audio seeks.
## CI
- Added unified `test` job replacing separate per-package test jobs
- Added root `test` script: `bun run --filter '*' test`
- New packages with test scripts are automatically included
- Added happy-dom for player DOM tests
## Tests
- 10 new tests for parent-frame media: preloading, play, pause,
seek, muted/rate sync, cleanup, attribute changes
- All 21 player tests pass
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(shader-transitions): pass CI when no test files exist
Add --passWithNoTests to vitest run so the unified test job
doesn't fail on packages that have a test script but no test
files yet.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): update tests for new id field and GSAP lint rule
- normalize.test.ts: loadTranscript now assigns id fields (w0, w1, etc.)
to SRT/VTT results and empty string for words-json passthrough
- lintProject.test.ts: add GSAP CDN script to validHtml() fixture to
satisfy the missing_gsap_script lint rule added in core
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): add missing data-start/data-duration to validHtml fixture
The validHtml() test fixture was missing data-start and data-duration
attributes, triggering the root_composition_missing_data_start and
root_composition_missing_data_duration lint warnings.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(ci): fetch LFS objects for producer test job
Producer regression tests compare rendered output against reference MP4
files stored in git LFS. Without lfs: true, checkout fetches pointer
files instead of actual videos, causing "moov atom not found" errors.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* ci: remove redundant test-producer job
The regression workflow already runs the same 28 producer fixtures
in a Docker container with prod-matching Chrome/fonts/ffmpeg, sharded
across 8 parallel matrix jobs with 40-min timeouts. The CI test-producer
job was a duplicate that ran on bare runners with worse determinism
and a 15-min timeout too short for all fixtures.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## What
PR 5/17 of the catalog system rollout. Adds the `hyperframes add` verb for installing blocks and components from the registry into an existing project, plus the `hyperframes.json` project config that tells `add` which registry to use and where to drop files. Stacks on #255.
- **`packages/cli/src/commands/add.ts`** — new `hyperframes add <name>` command. Resolves an item, validates target paths, installs files in parallel, builds an include snippet, copies it to the clipboard. Exposes a testable `runAdd(opts)` function; the citty default wraps it with console output + exit handling
- **`packages/cli/src/utils/projectConfig.ts`** — read/write/normalize `hyperframes.json`. Tolerant to missing and partial configs
- **`packages/cli/src/utils/clipboard.ts`** — minimal cross-platform clipboard (pbcopy / clip.exe / wl-copy / xclip / xsel). Zero deps. Gracefully no-ops in headless environments
- **`packages/cli/src/commands/init.ts`** — write `hyperframes.json` during scaffold if not already present
- **`packages/cli/src/cli.ts`** + **`help.ts`** — register `add` under Getting Started (directly below `init`)
Design doc: [Hyperframes Catalog System](https://www.notion.so/heygen/Hyperframes-Catalog-System-Design-Plan-341449792c69813f899dcd53b4c0383a).
## UX
```bash
# Scaffold a project (now writes hyperframes.json too)
npx hyperframes init my-video --example blank
cd my-video
# Add a block — files land, snippet copied to clipboard
npx hyperframes add claude-code-window
# ✓ Added claude-code-window (hyperframes:block)
# compositions/claude-code-window.html
#
# Include snippet:
# <iframe src="compositions/claude-code-window.html" data-start="0" data-duration="6"></iframe>
#
# Copied to clipboard — paste into your host composition.
# Add a component effect
npx hyperframes add shader-wipe
# Headless / CI — no clipboard, JSON output for tooling
npx hyperframes add shader-wipe --no-clipboard --json
```
Running `hyperframes add warm-grain` (an example) errors clearly pointing to `init --example`.
## Docs (bundled in this PR per the tracker principle)
- `docs/packages/cli.mdx` — new `add` subsection under Commands (flags, examples, trigger rules) + new `hyperframes.json` section describing the config file shape
## Tests
- **`packages/cli/src/commands/add.test.ts`** — 11 tests:
- `remapTarget` / `buildSnippet` pure helpers (5 tests)
- `runAdd` integration against a mocked `fetch` registry: block install lands files + returns snippet, component install respects `paths.components` remap, example-typed names throw `AddError` with code `example-type`, unknown names throw `AddError` with code `unknown-item` (4 tests plus 2 covering block default path and non-default path preservation)
- **`packages/cli/src/utils/projectConfig.test.ts`** — 9 tests:
- Write/read round-trip, partial-config normalization, corrupt-file handling, absent-file fallback to defaults, custom paths preserved
- **CLI suite:** 92 passed (was 72 on #255, **+20**). Same 4 pre-existing failures unchanged
## Scope decisions
- **`init.ts` full port to new resolver deferred.** The original plan bundled a removal of the `packages/cli/src/templates/` compat shim. That's ~300 more lines and isn't required for `add` to work. The compat shim from #254 still functions; a separate cleanup PR handles it
- **No ajv runtime schema validation.** Manifests are trusted as schema-valid. Full validation lands when third-party registries arrive (PR 14/15). Path safety is still enforced by the installer's `assertSafeTarget` guard
- **Default project paths stay under `compositions/`.** Blocks → `compositions/<name>.html`; components → `compositions/components/<name>/<file>`. Users override via `hyperframes.json#paths`
## Breaking / migration
**None.** Pure additive — new command, new file types, no existing commands or flags change. `init.ts` now writes `hyperframes.json` but that's a new additional file, not a modification of existing output.
## Stacks on
#255 — base branch. When #255 merges, this rebases onto `main`.
## Next in stack
PR 6 — `feat(registry): seed block — claude-code-window`. First real registry item. Exercises the full `hyperframes add <name>` flow end-to-end against a committed item on `main`.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
## What
PR 4/17 of the catalog system rollout. **Single clean cut** — the old flag is gone, replaced by `--example`. Alias changes from `-t` to `-e`. Stacks on #254.
- Rename `--template` → `--example` (alias `-e`) on `hyperframes init`
- Accept `--template` as a recognized-but-errored flag so users get a clear rename hint instead of citty silently ignoring the flag and producing a blank project
- Update all user-visible strings that referenced "template" as a user-facing concept in the init flow (picker prompt, step comments, offline-fallback suggestion)
- New `init.test.ts` covering both the success case (`--example` scaffolds) and the error case (`--template` exits 1 with rename hint)
Design doc: [Hyperframes Catalog System](https://www.notion.so/heygen/Hyperframes-Catalog-System-Design-Plan-341449792c69813f899dcd53b4c0383a).
## ⚠️ Breaking change
`--template` is no longer accepted. Example:
```bash
# before
npx hyperframes init my-video --template warm-grain
# after
npx hyperframes init my-video --example warm-grain
```
Users who still type the old flag will see:
```
The --template flag was renamed to --example. Example:
npx hyperframes init my-video --example warm-grain
```
and the command exits with code 1. This is **user guidance, not backwards compat** — the old flag's behavior is fully gone.
## Docs (bundled per the tracker principle)
- `docs/templates.mdx` — every `--template` reference
- `docs/quickstart.mdx` — agent-mode and video-mode examples
- `docs/packages/cli.mdx` — prose, `--help` flag table, `-e` alias
- `packages/cli/src/docs/templates.md` — CLI-embedded help topic
- `README.md` and `CONTRIBUTING.md` — not affected (no flag references)
User-facing renames of the `templates.mdx` page title, nav entry, and URL route are deferred to PR 11 (catalog discoverability UX) as planned.
## Why
1. **"examples"** matches shadcn + Remotion convention for full-project scaffolds and frees the word "template" for future parameterization work (string templating, placeholder substitution)
2. Once `hyperframes add` lands in PR 5, "template" vs "block" vs "component" would be three subtly different concepts sharing one word — renaming the old one to "example" makes the taxonomy self-explaining
## How
- **citty silently ignores unknown flags.** Naively removing `--template` would cause `hyperframes init my-video --template warm-grain` to silently fall through and scaffold a blank project. So `--template` stays declared in the args schema, but its run handler immediately errors with a rename hint and exits 1
- **Internal names unchanged** — `templateId` local variables, `getStaticTemplateDir` function, `BUNDLED_TEMPLATES` constant. They're implementation details; their rename is scheduled for PR 5 when the compat shims in `packages/cli/src/templates/` are fully removed alongside the `init` refactor
## Test plan
- [x] `bun run test` in `packages/cli`: **72 passed** (was 70 on #254, +2 new `init.test.ts` cases). Same 4 pre-existing failures unchanged
- [x] **New unit tests** in `init.test.ts`:
- `--example blank` non-interactive: exits 0, writes `index.html` to the target dir
- `--template blank` non-interactive: exits non-zero, stderr contains the rename hint + corrected command line, target dir is **not** created
- [x] **Manual smoke:**
- `npx hyperframes init /tmp/x --example blank` → "Created /tmp/x/"
- `npx hyperframes init /tmp/y --template blank` → "The --template flag was renamed to --example..." exit=1
- [x] `bunx oxfmt --check` + `bunx oxlint` on changed files: clean
- [x] Pre-commit typecheck (core + studio): clean
## Incidental fix
Resolver test regression from PR 3's simplify follow-up: `loadAllItems`' warning-path test was still spying on `console.warn` after the `onWarn` callback refactor. Now uses the callback directly.
## Stacks on
#254 — base branch. When #254 merges, this rebases onto `main`.
## Next in stack
PR 5 — `feat(cli): add command + hyperframes.json`. The big UX PR where:
- `init.ts` gets fully ported to the new registry resolver
- Compat shims in `packages/cli/src/templates/` are removed
- Users gain the `add` verb for installing blocks and components into existing projects
- `hyperframes.json` project-config file lands
🤖 Generated with [Claude Code](https://claude.com/claude-code)
## What
PR 3/17 of the catalog system rollout. Introduces the registry resolver/installer abstraction. No UX change — `init --template` still works identically. Stacks on #253.
**New module: `packages/cli/src/registry/`**
- `remote.ts` — fetches manifests (`registry.json`, `registry-item.json`) and item files from a GitHub-hosted registry. 24h cache on manifests; item files stream straight to `destDir`
- `resolver.ts` — `listRegistryItems`, `loadAllItems` (parallel fetch for picker UX), `resolveItem` (single-item fetch with `Available:` error)
- `installer.ts` — `assertSafeTarget` (runtime path-traversal guard) + `installItem` (parallel file download with up-front validation; all-or-nothing semantics)
- `index.ts` — barrel
**Registry content:**
- `registry/registry.json` — top-level manifest in PR 1's `RegistryManifest` shape. 8 examples
- `registry/examples/<id>/registry-item.json` — per-item manifest for each existing example, generated from legacy `templates.json` + HTML data-attribute probing
- `registry/examples/templates.json` — **deleted**, replaced by the above
**Compat layer:**
- `packages/cli/src/templates/{remote,generators}.ts` — thin shims that delegate to `../registry/`, keeping `init.ts`'s existing imports stable. `init.ts` doesn't move to the new API until PR 5 where it's part of a larger UX pass
**Tooling:**
- `scripts/generate-registry-items.ts` — idempotent one-off generator for this PR, kept in-repo for future example additions (`--only <name>` flag)
Design doc: [Hyperframes Catalog System](https://www.notion.so/heygen/Hyperframes-Catalog-System-Design-Plan-341449792c69813f899dcd53b4c0383a). Tracker entry in local `hyperframes-catalog-plan.md`.
## Why
Every future PR (`hyperframes add`, seed blocks, seed components, custom registries) otherwise has to keep piling onto the ad-hoc fetch + `cpSync` pattern in the old `fetchRemoteTemplate`. The new module is the single place that understands the registry wire format and file layout. **This is also where PR 1's schema comes alive.**
## How
### Scope-trimmed from the plan
- **No transitive dependency resolution yet.** Examples have no deps today. `resolveItem` doesn't walk `registryDependencies`; PR 5 adds that when blocks/components need it.
- **No ajv schema validation yet.** TS types + runtime path-traversal guard are the only safety nets. Full JSON-Schema validation lands when the registry starts accepting third-party content (PR 14 / custom registries).
- **init.ts refactor deferred to PR 5.** Compat shims keep this PR small and reviewable. PR 5 rewrites init alongside adding the `add` command.
### Safety
- `assertSafeTarget` rejects absolute paths, `..` segments, Windows drive letters, and any target that `path.resolve` shows to escape `destDir`. Mirrors the PR 1 schema `pattern`/`not.anyOf` on `target`, but runs at install-time so a registry that bypasses schema validation still can't write outside the project
- Up-front validation in `installItem` means a malformed item fails **before** any file is written. Atomic-ish semantics: all files land or none do
### Caching
- 24h manifest cache lives at `~/.hyperframes/cache/` per existing convention, but now keyed by `<baseUrl>__<kind>__<name>.json` so PR 14 custom registries can coexist
## Test plan
- [x] `bun run test` in `packages/cli`: **70 passed** (was 57 on #253, +13). Same 4 pre-existing failures (SRT/VTT whisper normalizer + `lintProject` clean-project test) — identical to main. No regressions
- [x] **Resolver unit tests (8):** filter by type, parallel load with fail-safe, resolve-by-name with `Available:` error message, unreachable-registry handling
- [x] **Installer unit tests (5):** accepts simple relative paths, rejects `..` segments, rejects Unix absolute paths, rejects Windows drive letters, permits `.` and dotfile-like names
- [x] **Smoke test**: `hyperframes init /tmp/x --template blank` (bundled code path, unchanged) works end-to-end
- [x] `bunx oxfmt --check` + `bunx oxlint`: clean
- [x] Pre-commit typecheck (core + studio): clean. CLI typecheck has 2 pre-existing errors (`render.ts`, `studioServer.ts` — unrelated `"mov"` format issue on main)
- [ ] **Smoke test remote fetch (`--template warm-grain`)** — verifiable only post-merge; registry paths live on `main` after this PR lands
## Breaking / migration
**No end-user-visible UX change.** `init --template <name>` still works the same way. Internally, `templates.json` is gone and the CLI now reads `registry.json` + `registry-item.json` per example.
Installed CLIs on old versions (`hyperframes@0.1.0`–`0.3.0`) already broke at PR 2 merge (see #253 rollout note). The next CLI release after this lands (`0.3.1`+) is the full fix.
## Commits
1. `generate-registry-items.ts` + generated manifests + deleted `templates.json`
2. Resolver + installer + compat shims
3. Unit tests
(All squashed into one commit on this branch; see `git log feat/registry-resolver ^refactor/registry-examples-dir`.)
## Stacks on
#253 — base branch. When #253 merges, this rebases onto `main`.
## Next in stack
PR 4 — `feat(cli)!: rename --template to --example`. Single clean cut, no alias. Tiny PR (~150 lines) that mostly updates `init.ts`'s argument schema, help text, and docs. Depends on this PR so the new flag name can be applied against the refactored code path.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
## What
PR 2/17 of the catalog system rollout. **Physical directory rename.** Stacks on #252.
- `git mv templates/ registry/examples/` — all 8 example directories (`decision-tree`, `kinetic-type`, `nyt-graph`, `play-mode`, `product-promo`, `swiss-grid`, `vignelli`, `warm-grain`) plus `templates.json`
- `packages/cli/src/templates/remote.ts` — `TEMPLATES_DIR` constant from `"templates"` → `"registry/examples"`, exported for regression testing
- `scripts/generate-template-previews.ts` — `remoteTemplatesDir` resolved to the new path
- Comment updates in `packages/cli/src/templates/generators.ts` and `packages/cli/src/commands/init.ts`
- New regression test `packages/cli/src/templates/remote.test.ts` pinning the path constants so future reverts fail a test instead of silently breaking installed CLIs
Design doc: [Hyperframes Catalog System](https://www.notion.so/heygen/Hyperframes-Catalog-System-Design-Plan-341449792c69813f899dcd53b4c0383a).
## Why
The current `templates/` directory is a flat "things that scaffold projects" bucket. The catalog model splits content into three tiers: **examples** (full projects — what today's templates are), **blocks** (sub-compositions), and **components** (effect snippets). `registry/examples/` is the canonical home for what was previously at `templates/`, and this PR makes room for `registry/blocks/` and `registry/components/` in future PRs without top-level clutter.
## How
- `git mv` preserves file history — GitHub renders these as renames, not deletions + additions.
- Remote template fetch via giget reads `TEMPLATES_DIR`, so updating that one constant is sufficient for the CLI's remote code path.
- The CLI's **internal** `packages/cli/src/templates/` directory (which holds the `blank` and `_shared` bundled assets plus `generators.ts`/`remote.ts`) is a separate concept and is **not** touched here. Renaming that module belongs to PR 3 where the abstraction changes to a registry resolver.
- `templates.json` keeps its existing shape and location (now at `registry/examples/templates.json`). **PR 3 will transform it** to the new `registry.json` shape introduced in PR 1 and generate a per-item `registry-item.json` for each example. Leaving the shape change to PR 3 keeps this PR a pure physical move.
## ⚠️ Breaking change for previously-installed CLIs (`hyperframes@0.1.0` – `0.3.0`)
**What happens:** every published CLI version has `TEMPLATES_DIR = "templates"` baked in. After this PR lands on `main`, those CLIs will 404 on:
- `raw.githubusercontent.com/heygen-com/hyperframes/main/templates/templates.json` (manifest list) — caught silently in `listRemoteTemplates`, so the template picker falls back to showing only `blank`
- `github:heygen-com/hyperframes/templates/<id>#main` (giget download) — raises "Template downloaded but missing index.html"
**Decision: accept the break.** Hyperframes is pre-1.0 OSS with a small installed base; complex mitigations (dual-path fetch, redirect stubs, manifest-at-old-path with empty array) add permanent maintenance cost for a one-time rename.
**Rollout plan:**
1. Merge #252 (PR 1 — types & schemas) first
2. Merge this PR (#253)
3. Ship a patched CLI release (`hyperframes@0.3.1`) in the same work-day. Already-pinned old CLIs break on remote examples, but upgrading restores full functionality
4. Note the break in release notes + `CHANGELOG.md` under the `0.3.1` entry
Users still on an older CLI will see the failure only if they invoke `hyperframes init` with `--template <non-blank>`; `--template blank` (bundled) continues to work offline on every version.
## Test plan
- [x] `bun run test` in `packages/cli`: **57 passed** (was 55 on main, +2 regression tests for the path constants). Same 4 pre-existing failures (SRT/VTT whisper normalizer + `lintProject` clean-project test) — unchanged from main. No regressions
- [x] **Manual smoke test**: `hyperframes init /tmp/x --template blank` works (bundled code path, unchanged)
- [x] `bunx oxfmt --check` + `bunx oxlint`: clean
- [x] `bun run typecheck` (core + studio, pre-commit hook): clean
- [ ] **Manual smoke test for remote fetch (`--template warm-grain`)** — not verifiable locally before merge. Remote fetch resolves `github:heygen-com/hyperframes/registry/examples/<id>#main`, which doesn't exist until this PR lands. Will work on `main` immediately after merge.
## Breaking / migration
- Internal repo path changes only. `--template` CLI flag continues to accept the same template names.
- See "Breaking change for previously-installed CLIs" above — decision is to ship a simultaneous CLI release rather than add a compat shim.
## Commits
1. `d691bd1` — initial rename + CLI path constant update
2. `fc0c642` — review feedback: docstring fix, regression tests, clarifying comment in `init.ts`, export constants for testing
## Stacks on
#252 — base branch. When #252 merges, this rebases onto `main`.
## Next in stack
PR 3 — `feat(cli): registry resolver + installer`. Transforms `templates.json` to the new `registry.json` shape (from PR 1's schema), generates `registry-item.json` for every existing example, introduces `packages/cli/src/registry/{resolver,installer,remote}.ts`, renames the `packages/cli/src/templates/` CLI module, and refactors `init` to call through the new abstraction.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
* feat(core): add registry schema + TS types
PR 1/17 of the catalog system rollout. Foundation for a shadcn-style
registry with three item tiers: examples (full projects), blocks
(sub-compositions), components (effect snippets).
## What
- TS types: RegistryItem (discriminated union of ExampleItem/BlockItem/
ComponentItem), RegistryManifest, FileTarget, ItemType, FileType
- JSON Schemas: schemas/registry.json, schemas/registry-item.json
- Compile-time exhaustiveness asserts on ITEM_TYPES/FILE_TYPES so adding
to the TS union without updating the constant stops compiling
- Drift-guard test: schema enums must equal ITEM_TYPES/FILE_TYPES by
set-equality; exactly 2 distinct type enums in registry-item.json
- Public API via new ./registry export path plus re-exports from root;
schemas exposed via ./schemas/registry.json export for external tooling
## Why
- Every downstream PR (resolver, installer, hyperframes add, docs
codegen, CI previews, skill, catalog command) builds on these types
- Getting the shape right now avoids painful migrations later
## How
- Discriminated union enforces that components do not have dimensions
or duration and examples/blocks must have them (schema mirrors via
if/then/else on the type discriminant)
- target path pattern rejects .. segments, Unix absolute paths, and
Windows drive letters (defense-in-depth; CLI validates at runtime in
PR 3)
- name pattern requires alphanumeric start and end (no trailing hyphens)
- Optional metadata: version, author, license, deprecated, minCliVersion
- additionalProperties: false on nested objects (catches typos on
critical fields) but relaxed on top-level RegistryItem (allows
third-party custom metadata in PR 15 custom registries)
## Test plan
- [x] Unit tests: 11 new tests covering type guards, discriminant
narrowing, schema/TS drift guards, schema \$id sanity, optional
metadata acceptance, and compile-time checks (via @ts-expect-error)
- [x] bun run test in packages/core: 445 passed (was 434 on main,
+11 from this PR)
- [x] bunx oxfmt and bunx oxlint: clean
- [x] bun run typecheck: clean
- [ ] Manual testing: N/A (types + schemas only)
- [ ] Documentation updated: per-item doc pages land in PR 9 (codegen
from these manifests); guide updates in PR 10+
## Breaking / migration
None. Pure additive — new module, new export paths, no existing
surface touched.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(core): remove version field + hyperframes:demo file type
Address review feedback from Miguel:
- Remove `version` from RegistryItemBase + schema. Per shadcn model,
the registry is versioned by git tags, not per-item. The adversarial
review added it; the original design doc was correct.
- Remove `hyperframes:demo` from FileType union + FILE_TYPES constant
+ schema. Demo files exist on disk for the CI preview pipeline but
are NOT installed to user projects and should not appear in
registry-item.json files[]. Neither shadcn nor Remotion has a
dedicated demo file type — demos are just compositions.
- Add `required: ["type"]` to the if-condition in the schema's
allOf discriminant (Miguel's nit — makes the condition self-
contained)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(ci): add shader-transitions to Dockerfile.test
PR #251 added packages/shader-transitions/ to the workspace but didn't
update Dockerfile.test to COPY its package.json. This caused
`bun install --frozen-lockfile` to fail in the regression Docker build:
bun saw a lockfile referencing @hyperframes/shader-transitions but the
package.json wasn't present in the container, so it wanted to remove
the entry — triggering "lockfile had changes."
Verified: Docker build passes with `--no-cache` after this fix.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(quickstart): collapse prerequisites into expandable accordion
Wraps the Node.js and FFmpeg install instructions in an Accordion
component so the quickstart page is less verbose for returning users
who already have the dependencies installed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(quickstart): add prerequisite bullet list above accordion
Adds a concise bullet list (Node.js 22+, FFmpeg) above the expandable
install instructions so users can see at a glance what's needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(player): handle Infinity duration from runtime gracefully
When compositions have repeating animations without data-duration, the
runtime sends durationInFrames: Infinity. The player now ignores
non-finite duration values instead of displaying "Infinity:NaN" in the
controls. formatTime also returns "0:00" for non-finite inputs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(lint): add data-duration and Math.ceil overshoot rules
- Add root_composition_missing_data_duration warning when the root
composition element is missing data-duration, which causes the runtime
to infer Infinity for loop-inflated timelines.
- Add gsap_repeat_ceil_overshoot warning that catches
repeat: Math.ceil(d/c)-1 patterns which overshoot the intended
duration. Recommends Math.floor instead.
- Fix gsap_infinite_repeat fixHint to suggest Math.floor (not Math.ceil).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(player): wait for injected runtime before declaring ready
When the player auto-injects the runtime script (because the
composition has GSAP timelines but no runtime), it would immediately
declare ready on the next probe cycle — before the runtime script
finished loading from CDN. This caused play() to send a postMessage
that nobody received, making autoplay silently fail.
Now the probe waits for the runtime bridge (__hf or __player) to
appear before proceeding to the ready state.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The tts command was implemented (PR #201) but never added to the root-level
help display or documentation. This adds it to:
- help.ts GROUPS (AI & Integrations) so it appears in `hyperframes --help`
- docs/packages/cli.mdx with usage examples and flag reference
- CLAUDE.md "Adding CLI Commands" checklist: new steps 4-5 require adding
commands to help.ts groups and docs, preventing future omissions
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a FIFO semaphore to limit concurrent renders in the producer server,
preventing Chrome CPU contention that causes beginFrame failures.
- New Semaphore utility class (packages/producer/src/utils/semaphore.ts)
- Both blocking render and SSE renderStream handlers acquire/release the semaphore
- SSE stream sends a "queued" event when request must wait
- New GET /render/queue endpoint exposes active/queued render counts
- Configurable via HandlerOptions.maxConcurrentRenders or PRODUCER_MAX_CONCURRENT_RENDERS env var (default: 2)
- New --max-concurrent-renders CLI flag (1-10)
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): smart port selection with instance reuse
Replace the simple 10-port retry loop with best-in-class port handling:
- Multi-host port testing (127.0.0.1, 0.0.0.0, ::1, ::) catches ports
occupied by SSH forwarding or other interfaces invisible to localhost
- HTTP probe (/__hyperframes_config) detects existing HyperFrames
preview servers — reuses same-project instances instead of spawning
duplicates, skips different-project instances
- PID detection via lsof for actionable "Port N in use by PID X" logs
- Expanded scan range from 10 to 100 ports
- Added --force-new flag to bypass instance detection
- Async PID detection (execFile, no shell) and parallel host testing
Fixes the "10 ports are all in use" error that occurs when zombie
preview servers accumulate or devbox port forwarding occupies ports.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add --list and --kill-all flags to preview command
- `hyperframes preview --list` scans the port range and displays all
active HyperFrames preview servers with their project name, directory,
and PID
- `hyperframes preview --kill-all` kills all active preview servers
- Port scanning uses parallel batched probes (20 at a time) for speed
Gives users visibility into zombie preview servers and a one-command
way to clean them up.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: apply HyperFrames design system to Mintlify theme
Update docs config and add custom CSS to match the HyperFrames brand:
- Switch theme from mint to maple, replace cyan palette with warm neutrals
- Add Inter (body/headings) and IBM Plex Mono (code) fonts
- Add custom.css with full light/dark mode CSS variables
- Default to light mode appearance
- Replace box-shadow hover effects with border-color (flat aesthetic)
- Add DESIGN.md to repo root as design system reference
- Fix docs CI to also trigger on DOCS_GUIDELINES.md pushes to main
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: replace HeyGen logo with HyperFrames text wordmark
Replace 41KB HeyGen SVG logos with lightweight (~400B) text-based SVGs
rendering "HyperFrames" in Inter semibold with tight tracking, matching
the wordmark style on hyperframes.heygen.com.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: use ABC Solar Display font for logo wordmark
Match the exact font rendering from hyperframes.heygen.com:
- Load ABC Solar Display Bold from HeyGen static assets CDN
- SVGs use 15.2px/600w/-0.15 letter-spacing (matches computed styles)
- Dark mode fill matches rgb(240,240,240) from the website
- Add @font-face in custom.css for site-wide availability
- Fix lefthook: remove css from oxfmt glob (oxfmt doesn't support CSS)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: convert logo SVGs to outlined paths
SVG <text> elements don't render custom fonts when loaded as <img>
(browser security restriction). Convert the ABC Solar Display glyphs
to SVG paths extracted from the font outlines — renders identically
everywhere with zero font dependency. Remove @font-face for the
display font from custom.css since it's no longer needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: constrain logo height to match website sizing
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Revert "docs: constrain logo height to match website sizing"
This reverts commit 89e9cb598e.
* docs: switch body font from Inter to TT Norms Pro
Use TT Norms Pro (from HeyGen static assets CDN) to match
hyperframes.heygen.com. Loads weights 400-700 via @font-face
with Inter as fallback.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Description
Adds proper BT.709 color space metadata and full→limited range conversion to H.264/H.265 encoding. Chrome captures frames in full-range sRGB (BT.709 primaries), but without explicit color tagging, players guess the wrong color space and range — causing color shifts across iOS/Android/desktop and crushed dark values that compound the gradient banding issue fixed in #222.
**What changed:**
| Setting | Before | After |
|---------|--------|-------|
| `color_space` | `bt470bg` (guessed) | `bt709` (explicit) |
| `color_primaries` | `unknown` | `bt709` |
| `color_transfer` | `unknown` | `bt709` |
| `color_range` | `pc` (full, wrong for H.264) | `tv` (limited, correct) |
| `time_base` | `1/15360` (varies by platform) | `1/90000` (fixed) |
**Approach:**
- BT.709 VUI params embedded via x264-params/x265-params (`colorprim=bt709:transfer=bt709:colormatrix=bt709`) — ensures the bitstream itself carries color info
- FFmpeg-level metadata flags (`-colorspace:v bt709`, etc.) — belt-and-suspenders
- `scale=in_range=pc:out_range=tv` filter converts Chrome's full-range output to TV/limited range
- VAAPI path chains the range filter with existing `format=nv12,hwupload`
- `-video_track_timescale 90000` for consistent cross-platform A/V timing (same as Remotion)
- VP9 and ProRes encoding unaffected
## Testing
- Verified via ffprobe: all 5 color metadata fields now correct
- Directly tested FFmpeg args produce expected output
- 40 engine tests pass (8 new: color metadata h264/h265, range filter CPU, VAAPI filter chain, GPU skip, VP9 skip, timescale)
- Builds cleanly, lint + format pass
Add aq-mode=3 (auto-variance adaptive quantization) to CPU H.264/H.265
encoding. This redistributes bits from bright/textured areas to dark flat
areas where color banding is most visible in 8-bit yuv420p output.
- standard/high presets: aq-mode=3 + aq-strength=0.8 + deblock=1,1
- draft (ultrafast): aq-mode=3 only (deblock too slow for ultrafast)
- GPU and VP9 encoders unaffected (have their own AQ implementations)
Adds 6 regression tests verifying the params are emitted correctly.
Fixes color banding on dark gradients (eval issue #3, prompts 3,5,10,14).
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(skills): consolidate 15 skills into 3 for better trigger reliability
Merge 9 GSAP skills (core, timeline, scrolltrigger, plugins, utils, react,
frameworks, performance, effects) and 6 HyperFrames skills (compose, captions,
tts, audio-reactive, marker-highlight, cli) into 3 consolidated skills:
- `gsap` — core API + timelines + performance in SKILL.md; scrolltrigger,
plugins, utils, react, frameworks, effects in references/
- `hyperframes` — composition authoring rules in SKILL.md; captions, tts,
audio-reactive, marker-highlight in references/
- `hyperframes-cli` — CLI commands (init, lint, preview, render, etc.)
Why: With 15 separate skills, agents must correctly trigger the right subset
for any task. "Create an animated video with captions" needed 6+ skills to
fire — each with ~90% trigger accuracy means ~53% chance of getting all of
them. With 3 skills, that same task needs just `hyperframes` + `gsap` (~90%
both fire). Progressive disclosure still works via references/ files loaded
on demand.
Also fixes: CLAUDE.md referenced `window.__GSAP_TIMELINE` (incorrect) —
corrected to `window.__timelines`.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add --skip-skills flag to init command
Allow skipping the AI coding skills installation prompt during
`hyperframes init` with `--skip-skills`. Useful when skills are
already installed or when the user wants to scaffold without them.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(skills): address code review feedback on consolidation
Restore content lost during over-compression:
- captions: fix overflow to `visible` (not hidden — clips glow effects),
add container pattern warning, scale headroom formula, and self-lint
placement guidance
- audio-reactive: restore sampling frequency pattern (per-frame tl.call
loop vs single tween) and textShadow-on-container gotcha
- effects/typewriter: restore word rotation, appending words, spacing
with static text, and multi-line cursor handoff patterns
- effects/audio-visualizer: restore spatial mapping conventions, fetch vs
inline loading, WebGL/DOM rendering approaches, and canvas layering
- hyperframes-cli: restore --strict-all flag in render flags table
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): update build:copy and template for consolidated skill names
- build:copy: reference skills/hyperframes, skills/hyperframes-cli,
skills/gsap instead of the old 15 skill directory names
- _shared/CLAUDE.md template: update skill table to consolidated names
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace the static "Tip" message at the end of `hyperframes init` with an
interactive prompt that offers to install AI coding skills. When the user
accepts, the skills command runs `npx skills add` with `--all` and
`stdio: "inherit"` so the native installer output is visible.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move per-command examples from the centralized `help.ts` record into
each command file as `export const examples: Example[]`. help.ts now
dynamically imports them at --help time.
This means adding a new command and its examples happens in one file
instead of two, reducing the chance of forgetting examples.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add `tts` command for local text-to-speech via Kokoro-82M
Adds `hyperframes tts` — generate speech audio locally using Kokoro-82M
(ONNX), no API key needed. Mirrors the transcribe command architecture.
- New command: `hyperframes tts "text" --voice af_heart --output speech.wav`
- 54 voices across 8 languages, ~5x realtime on CPU
- Auto-downloads model (~311 MB) + voices (~27 MB) to ~/.cache/hyperframes/tts/
- Requires Python 3.8+ with kokoro-onnx installed
- Extracted shared `downloadFile` utility from whisper/manager.ts with
atomic .tmp→rename to prevent partial download corruption
- Added hyperframes-tts skill with voice selection guide
- Updated CLAUDE.md with TTS docs, voice table, and skill reference
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(tts): improve skill per skill-creator guidelines
- Move trigger info from body to frontmatter description
- Remove `trigger` field (not a valid frontmatter field)
- Remove CLI flag docs Claude can derive from --help
- Remove redundant voice tables (keep content-to-voice mapping)
- Fix composition audio example to use actual <audio> element pattern
- Keep non-obvious workflows: TTS+transcribe for captions, long scripts
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(tts): add guidance for using external TTS sources
Help users understand when to use cloud TTS (voice cloning, broader
languages, higher quality) vs the built-in Kokoro model, and how
external audio integrates into the same composition workflow.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(tts): prioritize HeyGen API as recommended cloud TTS
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(tts): remove external TTS section for now
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(tts): set required: false on input arg so --list works standalone
Citty treats positional args as required by default unless explicitly
set to required: false. Without this, `hyperframes tts --list` fails
with "Missing required positional argument".
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(tts): add --help examples and fix required:false for --list
Add examples section to `tts --help` matching the pattern from other
commands (transcribe, render, etc.). Fix citty positional arg requiring
explicit `required: false` for --list to work standalone.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: add CLI command checklist to CLAUDE.md
Ensure new commands always get --help examples in help.ts.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): improve --help with grouped commands and per-command examples
Replaces citty's flat COMMANDS list with kubectl-style grouped categories
and adds examples to every subcommand. Root help now groups 14 commands
into 5 categories (Getting Started, Project, Tooling, AI & Integrations,
Settings). Per-command --help now appends a formatted Examples section
with practical usage patterns.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(cli): use shared colors, lazy-load help, fix description drift
- Replace hand-rolled ANSI helpers with existing ui/colors.ts (fixes
non-conformant NO_COLOR handling)
- Add cyan and gray to shared color module
- Lazy-load help.ts via dynamic import to avoid allocating help data
on non-help invocations
- Fix description drift: benchmark and transcribe descriptions now
match their command meta.description
- Unify tuple order: ROOT_EXAMPLES now uses [comment, command] to
match COMMAND_EXAMPLES
- Remove redundant comment restating type annotation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): resolve typecheck error in showUsage generic signature
The lazy-load wrapper needs to cast CommandDef<T> to CommandDef when
forwarding to the help module, since TypeScript's generic variance
makes the direct assignment incompatible.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* chore(cli): remove kubectl references from help comments
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add remote template fetching via giget
* fix: update remote.ts to use templates/ instead of examples/
* fix(cli): validate template ID against manifest before downloading
Fails fast with available template list instead of downloading
an empty directory for nonexistent templates.
* refactor(cli): simplify to single --template flag with dynamic validation
- Remove --example flag (--template handles bundled + remote)
- Remove static ALL_TEMPLATE_IDS list (validates against GitHub manifest)
- No CLI release needed to add new templates — just add to templates/ and templates.json
- scaffoldProject auto-detects bundled vs remote
* chore: update lockfiles for giget dependency
* fix(cli): remove undefined isAudioOnly reference
* refactor(cli): move templates to examples/ for remote fetching
* refactor(cli): rename examples/ to templates/ for clarity
Follows Remix/Vite convention — these are init scaffolds, not general examples.
* feat(docs): add template gallery page with visual previews
* fix(docs): remove invalid MDX heading anchors
* chore: retrigger CI
* feat(docs): merge gallery into templates page with hover-to-play video previews
- Consolidated gallery.mdx and templates.mdx into single templates.mdx
- Moved templates page to Getting Started section
- Added MP4 video previews rendered by hyperframes (hover to play)
- Custom JS for hover-to-play behavior (Mintlify strips JSX event handlers)
- 2-column grid for landscape, 3-column for portrait
- Remotion-style cards with gradient overlay labels
* fix(docs): update broken links after templates page move
* ci(regression): remove scripts/ from regression trigger paths
scripts/ contains dev utilities (lint, versioning, preview generation)
that don't affect the rendering engine.
## What
Add a script that uses `@hyperframes/producer` to render PNG thumbnails of each built-in template. Output goes to `docs/images/templates/`.
## Why
User feedback: "I would have loved more visual examples of what's actually possible. I had to scaffold every template just to see what they look like." This is the foundation for a visual template gallery in the docs.
## How
- New `scripts/generate-template-previews.ts` uses the producer's `createFileServer` + `createCaptureSession` + `captureFrame` APIs — hyperframes renders its own templates
- Patches out `__VIDEO_SRC__` placeholders (same logic as `init.ts`) so templates render without a video file
- Captures a frame at t=2s for each template (skips `blank` — it's just empty scaffolding)
- Handles varying dimensions (vignelli is 1080x1920 portrait)
- Adds `pnpm generate:previews` npm script
## Test plan
- [x] `pnpm generate:previews` generates 8 PNGs in `docs/images/templates/`
- [x] Each PNG is visually correct (verified by viewing)
- [x] `--only <template>` flag works for single template generation
- Replace hardcoded template list in build:copy with `cp -r src/templates/*`
so new templates are included automatically (kinetic-type, decision-tree,
product-promo, nyt-graph were missing from published package)
- Fix captions SKILL.md: reword `!` and `>` in inline backticks that
triggered Claude Code's bash permission checker
- Add scripts/lint-skills.ts to catch shell-unsafe patterns in SKILL.md
files (runs as part of `bun run lint` in CI)
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli,core): add standalone transcribe command, transcript normalization, and caption lint rules
- Add `hyperframes transcribe` command for transcribing audio/video and importing
existing transcripts (SRT, VTT, OpenAI Whisper API JSON, whisper.cpp JSON)
- Add transcript format normalizer (normalize.ts) with auto-detection and
conversion to standard [{text, start, end}] word arrays
- Upgrade default whisper model from base.en to small.en for better accuracy
- Add --model and --language flags to both `transcribe` and `init` commands
- Extract shared patchCaptionHtml() to eliminate duplication between init.ts
and transcribe.ts (init.ts reduced by ~55 lines)
- Add 3 caption lint rules: caption_exit_missing_hard_kill,
caption_text_overflow_risk, caption_container_relative_position
- Update captions skill with model guide, format docs, music guidance,
text overflow prevention, caption exit guarantee pattern
- Expand captions skill trigger to cover lyrics, karaoke, lyric videos
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs(cli): add transcribe command and --model/--language flags to CLI docs
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): fix blank template lint issues
- blank/index.html: remove data-start from video (was nested in timed parent),
add class="clip" for initial hidden state
- blank/captions.html: add max-width + overflow:hidden to prevent text clipping,
add tl.set hard kill after exit tween to prevent stuck captions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: add lint-after-edit rule to repo and project CLAUDE.md
Agents must run `npx hyperframes lint` after editing compositions.
Also expand captions skill description in project template.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* style: format _shared/CLAUDE.md
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The --tag flag was optional, which led to v0.1.11 and v0.1.12 being
bumped without tags — skipping npm publish entirely. Invert the default:
always commit + tag, with --no-tag as the escape hatch.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add npm version and MIT license badges (matching React/Remotion/Next.js conventions)
- Remove non-standard Docs badge
- Move Documentation section from bottom of README to right after Quick Start
for better discoverability (follows patterns from React, Remotion, Vite)
- Link to quickstart at hyperframes.heygen.com/quickstart
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add system metrics to telemetry and expand doctor command
Enrich render telemetry with device/environment metadata (CPU, memory,
OS, Docker/CI/WSL detection) following patterns from Next.js and
Turborepo. Add speed_ratio (render time / composition duration),
per-frame capture timing, and resource usage to render events.
Expand the doctor command with CPU, memory, disk, /dev/shm, and
environment checks to help debug rendering issues on user machines.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): invert speed_ratio to match experiment-framework convention
composition_duration / render_time — higher is better, >1 means faster
than realtime. Matches magic_edit.render.speed_ratio in experiment-framework.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): wire errorMessage into render error telemetry
Address review feedback — the errorMessage field was declared in the
trackRenderError interface but never populated.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add render telemetry to embedded studio server
Track render_complete and render_error from the studio's render API
endpoint (hyperframes dev). Uses dynamic imports so telemetry is
resolved at call time within the CLI package — no telemetry coupling
added to @hyperframes/studio or @hyperframes/producer.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Namespace skill names with `hyperframes-` prefix for clearer identity in
OSS contexts where users may have other skills installed.
Updates skill directories, SKILL.md frontmatter, CLAUDE.md, README.md,
CLI build script, init command, and project template.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
James Russo