- 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>
## Summary
- Respect `--json` flag on all lint exit paths so agents always get machine-readable output
- Separate `infoCount` from `warningCount` in linter results (was conflated)
- Display `info` vs `warning` severity distinctly in lint output
* fix(lint): detect GSAP animations targeting clip elements (tab crash)
The runtime manages clip visibility via inline styles. When GSAP also
writes inline styles on the same element, both systems trigger style
recalculations every frame, creating a runaway loop that crashes the
browser tab.
New rule gsap_animates_clip_element (error severity):
- Builds map of all elements with class="clip" (by id and class)
- Checks if any GSAP selector resolves to a clip element
- Nested selectors like "#overlay .title" are correctly ignored
- Merged into existing GSAP script loop (no redundant parsing)
* fix: remove non-null assertions and add missing test coverage
- Replace `!` assertions with optional chaining in lint.ts and tests
- Add shouldBlockRender tests for --strict-all without --strict
- Add clip element test for class-only detection (no id)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: use optional chaining for array access in lintProject tests
TypeScript's strict mode flags array indexing as possibly undefined.
Use optional chaining and fallbacks instead of non-null assertions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
- **Lint rules** catch media elements missing `id` (renderer silently skips them), missing `src`, `preload="none"` (blocks renderer), and video nested in timed divs (freezes playback). Upgraded `video_nested_in_timed_element` from warning to error.
- **Compiler** strips `preload="none"` from media during compilation. Runs parallel, cached keyframe interval analysis via ffprobe — warns on sparse keyframes (>2s) that cause seek failures and audio/video desync. Suggested ffmpeg command preserves audio (`-c:a copy`).
- **Pre-render lint** lints `index.html` + all `compositions/*.html` sub-compositions before render via shared `lintProject()` helper. Warns by default; `--strict` blocks on errors, `--strict-all` blocks on errors + warnings.
- **Render orchestrator** logs a hint to retry with `--workers 1` when parallel capture times out on video-heavy compositions.
- **Refactor**: extracted `runFfprobe()` + `parseProbeJson()` helpers to deduplicate ~80 lines of spawn boilerplate across 3 ffprobe functions. Extracted `shouldBlockRender()` so strict flag tests exercise production code. Shared `lintProject()` used by both `lint` and `render` commands.
## Context
Discovered during a real composition build session where:
1. `<audio>` without `id` rendered silently (preview worked fine because runtime queries `[data-start]`, but renderer queries `[id][src]`)
2. `<video>` inside timed `<div>` froze on first frame
3. `preload="none"` caused 45s renderer timeout
4. YouTube clips with sparse keyframes from `yt-dlp --download-sections` caused audio/video desync
5. Parallel workers timed out on video-heavy compositions
## Test plan
- [x] Core: 365/365 tests passing (5 new lint tests)
- [x] Engine: 24/24 tests passing
- [x] CLI: 14/14 tests passing (7 lintProject + 7 shouldBlockRender)
- [x] Lint + format hooks pass
- [ ] Manual: create a composition with `<audio data-start="0" src="test.wav">` (no id) — verify `npx hyperframes lint` catches it
- [ ] Manual: run `npx hyperframes render --strict` with lint errors — verify it blocks
- [ ] Manual: run `npx hyperframes render --strict-all` with lint warnings — verify it blocks
## Summary
- `hyperframes lint` now scans ALL HTML files in the project (not just index.html)
- Includes composition files in `compositions/` directory
- `--json` output includes `file` field and `filesScanned` count
- Exit code 1 on errors for CI integration
## Examples
```bash
# Human-friendly output
hyperframes lint
# Agent/CI-friendly JSON
hyperframes lint --json
```
## Test plan
- [ ] `hyperframes lint` on a project with compositions → shows findings from all files
- [ ] `hyperframes lint --json` → outputs valid JSON with all findings
- [ ] Exit code 1 when errors found, 0 when clean
## Summary
- When running `hyperframes dev .` inside a symlinked directory, the project name showed the resolved target name instead of the visible directory name
- Now uses `$PWD` to preserve the user-facing name
- Added `projectName` option to `StudioServerOptions` for explicit override
## Test plan
- [ ] `ln -s /path/to/project my-project && cd my-project && hyperframes dev .` → should show "my-project"
- [ ] `hyperframes dev /path/to/project` → should show "project" (basename of path)
## Summary
Extracts all studio API routes into a shared Hono-based module at `@hyperframes/core/studio-api`.
### Architecture
- **`StudioApiAdapter` interface** — consumers inject host-specific behavior (project resolution, bundling, rendering, thumbnails)
- **Shared route modules**: projects, files, preview, lint, render, thumbnail
- **Shared helpers**: `isSafePath`, `walkDir`, `getMimeType`, `buildSubCompositionHtml`
### What this PR does
- Creates the shared module with all API routes extracted from both `vite.config.ts` and `studioServer.ts`
- Both consumers will be refactored in follow-up commits to mount this module with their own adapter
### What stays in each consumer
- **Vite**: SSR module loading, Puppeteer thumbnails, file watcher + HMR, producer HTTP proxy, multi-project scanning
- **CLI**: in-process `executeRenderJob`, local runtime serving, browser management, SPA static file serving
### Follow-up needed
- [ ] Refactor `packages/studio/vite.config.ts` to use `createStudioApi(adapter)` via `@hono/node-server`'s `getRequestListener`
- [ ] Refactor `packages/cli/src/server/studioServer.ts` to use `createStudioApi(adapter)`
- [ ] Add `./studio-api` export path to `packages/core/package.json`
- [ ] Add `hono` as peer dependency of `@hyperframes/core`
## Test plan
- [ ] Verify shared module compiles without type errors
- [ ] After consumer refactoring: all studio features work identically via both vite dev and CLI embedded servers
🤖 Generated with [Claude Code](https://claude.com/claude-code)
## Summary
- CLI render: use timestamped filenames (`project_date_time.ext`) matching the studio's naming convention, preventing overwrites of previous renders
- studioServer: read `fps`/`quality`/`format` from POST body instead of hardcoding `fps:30`/`quality:standard`/`mp4`
- studioServer: use timestamped job IDs matching the studio pattern
- studioServer: fix download endpoint to serve correct content-type for WebM
## Test plan
- [x] `hyperframes render --format webm` outputs timestamped WebM file
- [x] `hyperframes render` outputs timestamped MP4 (no overwrite)
- [x] Studio embedded server (`hyperframes dev`) renders with correct format when selected in UI
- [x] Download endpoint serves correct MIME type for WebM renders
* 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>
## What
Adds 4 new composition templates and fixes structural issues across all templates.
### New templates
- **decision-tree** — animated flowchart with branching paths
- **kinetic-type** — bold kinetic typography promo
- **product-promo** — multi-scene product showcase with SVG assets (3 scenes)
- **nyt-graph** — animated data chart in NYT print editorial style
### Fixes across all templates
- GSAP updated from 3.12.2 → 3.14.2 (all templates, including warm-grain, swiss-grid, vignelli, play-mode)
- New templates restructured with proper root wrapper div, `data-duration`, sub-composition refs with `data-composition-id` / `data-width` / `data-height`
- nyt-chart: replaced `${compId}` template literal variables with hardcoded `"nyt-chart"` string — cheerio's css-what parser crashes on template literals during bundling, causing silent fallback to raw HTML without runtime injection
- nyt-chart: added DOM readiness retry for dynamically created SVG elements
- kinetic-type: removed external S3 audio URL
- All templates: GSAP script loaded in `<head>` before any scripts reference it
## Why
The new templates expand the range of content types available via `hyperframes init`. The fixes ensure all templates work correctly in the studio preview (bundler inlines sub-compositions and injects the runtime).
## Test plan
- [x] All 4 new templates render in studio preview
- [x] nyt-graph chart animates bars, line, and labels on playback
- [x] Existing templates unaffected (GSAP version bump is backwards compatible)
- [x] `hyperframes lint` passes on all templates
- [x] `generators.ts` updated with new template IDs
## Summary
- Adds `process.exit(0)` after render completes in the CLI `render` command
- Fixes the process hanging indefinitely after `npx hyperframes render --output video.mp4`
- Root cause: Node.js `fetch()` keep-alive pool (from update checks and telemetry) keeps TCP connections open, preventing the event loop from draining
- Telemetry is preserved — the `exit` handler in `cli.ts` calls `flushSync()` which spawns a detached child process
## Test plan
- [x] Run `npx hyperframes render --output test.mp4` and verify the process exits after completion
- [x] Verify telemetry events are still sent (check PostHog)
- Init "next steps" now leads with opening in an AI agent as step 1,
followed by dev preview and render as steps 2 and 3
- Interactive init shows a tip about AI agent workflow after scaffolding
- Dev command output hints that compositions can be edited with an
AI agent and changes reload automatically in the studio
The shared CLAUDE.md in scaffolded projects mentioned
hyperframes.heygen.com alongside doc topic names, causing AI agents
to guess incorrect URLs like /compositions instead of using the local
`hyperframes docs` CLI command or the correct /concepts/compositions
path. Now explicitly directs agents to use the CLI for quick reference
and llms.txt for full doc URL discovery.
- Interactive wizard is now the default when running in a TTY terminal
- Non-interactive mode (for CI/agents) uses --non-interactive flag
- Template defaults to "blank" when not specified in non-interactive mode
- --template flag skips the template picker prompt in interactive mode
- Replaces --human-friendly flag with --non-interactive (inverted logic)
npm's sigstore provenance verification requires package.json
repository.url to match the GitHub repo. Without it, publish
fails with E422 "expected to match https://github.com/heygen-com/hyperframes".
- Compute CPU_CORE_COUNT once at module level instead of calling cpus()
multiple times
- Make RenderOptions.workers required (number, not optional) — the
caller resolves the default, callees don't re-derive it
- Remove redundant existsSync guard before mkdirSync({recursive})
- Trim oversized JSDoc on defaultWorkerCount
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace the hardcoded default of 4 workers with a CPU-aware heuristic:
half of available CPU cores, capped at 4. Each worker spawns a separate
Chrome browser process (~256MB RAM each), so the previous default of 4
caused resource contention on smaller machines.
The new defaults:
2-core laptop → 1 worker
4-core laptop → 2 workers
8-core desktop → 4 workers
16-core server → 4 workers (capped)
Also adds --workers auto flag support, improves help text to explain
what workers do, and adds a Workers section to the rendering docs with
guidance on when to increase or decrease parallelism.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Extract `resolveAssetDir()` helper to eliminate copy-paste across
getStaticTemplateDir, getSharedTemplateDir, getBundledSkillsDir
- Remove `counted` boolean in fallbackInstall() — collect installed
skills from first target explicitly, then copy to remaining targets
- Consolidate duplicate `installed.length > 0` check in runInstall()
into a single early return
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace custom git clone + copy logic with `npx skills add` from the
vercel-labs/skills ecosystem. This is the standard used by GSAP,
Remotion, and other skill providers.
- Primary path: `npx skills add heygen-com/hyperframes` and
`npx skills add greensock/gsap-skills` with `-g -y -a <agent>` flags
- Git clone + copy kept as fallback if npx is unavailable
- Added `skillsAgent` field to targets mapping CLI flags to the correct
agent names (e.g., gemini → gemini-cli)
- Removes ~80 lines of custom git/copy infrastructure
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New HyperFrames projects created via `hyperframes init` now include:
- CLAUDE.md + AGENTS.md — teaches AI agents about skills, commands,
project structure, and framework rules (class="clip", timeline
registration, determinism). Agents know to invoke /compose-video
before writing compositions.
- .claude/skills/{compose-video,captions} — project-level skills for
immediate availability in the current agent session (global skills
require a session restart to discover).
- Updated next-steps output with `hyperframes docs <topic>` and a
link to hyperframes.heygen.com.
- Updated README with "AI Agent Skills" section documenting
`npx hyperframes skills` and `npx skills add` install paths.
- Repo-level CLAUDE.md for framework contributors.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The blank template was a bare `<div>` fragment without `<!DOCTYPE>`,
`<head>`, or `<body>`. This caused:
- Blank preview (bundler/runtime can't initialize from a fragment)
- "Failed to run lint" (parsing errors on the malformed document)
All other templates (swiss-grid, vignelli, warm-grain, play-mode) are
proper HTML documents — blank was the only outlier.
Also improves lint error reporting to show the actual error message
instead of the generic "Failed to run lint."
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The SIGINT handlers that were meant to do graceful cleanup (close server,
remove symlinks) override Node's default exit behavior. If the cleanup
hangs (e.g. server.close() blocked by open connections), the process is
stuck and Ctrl+C does nothing.
Fix: don't intercept SIGINT at all. Node's default behavior exits the
process immediately on Ctrl+C. The OS reclaims the port and file handles.
Use process.on("exit") for best-effort symlink cleanup instead.
Also fixes running the CLI via `tsx` in dev mode — __CLI_VERSION__ is a
tsup build-time define that crashes at runtime without a fallback.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The embedded dev server (hyperframes dev) silently reported success even
when the requested port was already in use (e.g., by Cursor IDE on 3002).
This happened because the old isPortAvailable() probe had a TOCTOU race
and findAvailablePort() silently fell back to the original port on exhaustion.
Replace with serveWithPortFallback() that binds the real Hono server
directly via createAdaptorServer + manual listen(), retrying on EADDRINUSE.
Shows a yellow warning when auto-incrementing and a clear error when all
ports are exhausted.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Engine: document getEncoderPreset() for MP4/WebM, VP9 alpha flags,
Opus audio in mux step
- Producer: document format field in RenderConfig, WebM usage example,
pipeline steps updated for WebM
- CLI: add render command examples in --help (including WebM overlay)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Support rendering compositions with transparent backgrounds via
`--format webm`. VP9+alpha is the standard format for overlayable
video (captions, lower thirds, overlays).
Changes by layer:
- CLI: `--format mp4|webm` flag on render command
- Producer: threads format through RenderConfig, switches to PNG
capture and VP9 encoding when webm
- Engine: getEncoderPreset() returns VP9 config with yuva420p;
transparent page background via CDP when capturing PNG;
mux uses Opus audio for WebM; VP9 flags from production:
-row-mt 1, -auto-alt-ref 0, alpha_mode=1 metadata
- Frame capture: Emulation.setDefaultBackgroundColorOverride a=0
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Document the three dev server modes (embedded/local studio/monorepo)
- Add --port flag to dev command
- Document _meta envelope on all --json commands
- Document upgrade --check --json for agent consumption
- Document passive update notices and HYPERFRAMES_NO_UPDATE_CHECK
- Update doctor output example with Version check row
- Fix README default port from 3000 to 3002
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- New `updateCheck.ts` utility: cached npm registry check (24h TTL),
sync `getUpdateMeta()` for _meta envelope, `printUpdateNotice()` for
passive stderr banner
- `upgrade --check --json`: machine-readable version check for AI agents
Returns { current, latest, updateAvailable }
- `_meta` envelope on all --json commands (info, lint, benchmark,
compositions): includes version, latestVersion, updateAvailable
- `doctor` shows version check as first row
- Passive update notice on stderr after command completes (skipped in
CI, non-TTY, --json, --quiet)
- Background check fires on startup (non-blocking, populates cache)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
James