* perf(engine): a remote source's tail decodes run many at a time, not one per half-core
* perf(producer): the plan logs where its extract stage spends its time
* perf(engine): a window that ends inside its source takes its frame count without a decode
* test(engine): pin the tail-decode slot test to the windows that decode
* style(engine): format the tail-probe window check
* test(engine): deferred counts match extraction on a B-frame NTSC source; VFR is not deferred
* style(engine): format the tail-probe tests
* perf(aws-lambda): a site stores each video's ffprobe JSON, and the plan reads it instead of probing
* test(aws-lambda): a site stores and hands out each video's probe
* style(aws-lambda): format the probe-facts upload and tests
* perf(engine): a site's stored probe also holds the final-frame timestamp the plan would probe for
* test(engine): drop an unused import
* test(aws-lambda): the stored probe holds the ffprobe JSON and the final frame
* test(aws-lambda): the stored probe holds the ffprobe JSON and the final frame
* perf(producer): a distributed chunk encodes while it captures
* perf(producer): the audio task logs where its mix spends its time
One line per audio stage: prepare, groups, mix, master, encode and the peak check, in ms.
* perf(engine): locked warmup ticks stop sleeping once the page is ready
* perf(engine): a window reaching a stored video's end takes its frame count from stored facts
* fix(producer): the cloud audio task logs its phases
* perf(engine): stored facts keep the last packet length so late-start windows skip the tail decode
* fix(engine): a probe spawns without waiting when no stored facts are declared
* chore(aws-lambda): shorten comments over the line limit
* fix(engine): warmup keeps pacing after a failed tick; stored final frames check ffprobe major
* test(engine): the recorded ffprobe major matches the build's own banner
* test(engine): the version-banner probes go through a named binary like the other spawns
* fix(producer): a render's page-error count includes what the capture workers saw
* fix(producer): a failed parallel capture still carries its workers' page errors
* fix(producer): every capture failure carries the page's errors, in every stage
* chore(engine): drop a comment the capture classifier's name already says
* fix(engine): renames ride out Windows file locks instead of failing the render
* fix(producer): the true-peak and segment-manifest renames ride out Windows file locks too
parseAudioElements now resolves end the way core's compileTimingAttrs does. A clip
with no length at all warns, naming the clip, instead of silently playing to the end of
its source.
* fix(engine): h265 renders no longer fail on hosts with more than 16 cores
libx265 rejects more than 16 frame threads, so the thread count taken from the
CPU quota is now capped at 16 for x265. x264 is unchanged.
* fix(engine): pass the encoder to the thread cap in the CFR re-encode
* perf(producer): the audio task probes each video for sound instead of extracting its frames
* test(producer): the audio task finds a video's sound from a probe, with no frames extracted
* perf(engine): the audio mix reads clips held at a URL many at a time
* style(engine): format the remote clip slot choice
* fix(producer): the audio task places a video's sound in the window extraction would give it
* style(producer): format the slot cases of the script-src audio test
* perf(engine): remote sources of a site keep their probe results for the whole render
* test(engine): probe a content-addressed URL once per render
* style(engine): format the audio probe split
A `<canvas layoutsubtree class="hf-vfx-src">` nested inside another capture
canvas blocks the renderer main thread in drawElementImage, so `check` and
`snapshot` time out with no output. Refuse either direction of the nest with a
reportVfxError and leave the chain unregistered.
Also scope the two host lookups to immediate children (`:scope >`) so a nested
host's src/out canvas is never bound to the outer host, and read the
`data-vfx-ref-visible` flag in resolveCaptureSource so a canvas reached as both
a host's own source and as a ref gets one consistent `visible`/keepBitmap.
Refs #4405
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* perf(aws-lambda): a render reads a video's audio from a small file stored beside it
A video's audio is interleaved with its pictures, so mixing it from the cloud read the whole video (2.5 GB for a 20-minute 4K film) at the Lambda's network speed. lambda sites create now stores each large video's audio packets as a separate object keyed by the video's hash; the audio task reads that when it exists.
* chore(producer): shorten the audio sidecar header comment
* fix(aws-lambda): no audio sidecar for a video whose audio starts late or that has several tracks
* test(producer): format the audio sidecar tests and end their ffprobe options with --
* fix(producer): a chunk with mixed capture modes retries whole in screenshot mode
* fix(producer): format the mixed capture mode change
* fix(engine): the BeginFrame start-up check waits for the first frame up to its deadline
* style(engine): format the probe tests
* fix(engine): drop a comment from the probe loop
* fix(engine): the video frame cache budget is shared by the capture workers
Each capture worker built its own frame cache of up to 1500 MB of base64 strings, all on the one Node heap, then serialised every frame body for the browser. With many workers and 4K png frames the heap ran out inside JSON.stringify and the render aborted. The configured budget is now split evenly across the workers, never under 64 MB each.
* chore(engine): format and trim the frame cache budget change
* chore(engine): format the frame cache budget test
* fix(producer): chunk capture workers share the frame cache budget too
The distributed chunk path built its injectors with no options, so each worker kept the full 1500 MB cache whatever the config said. The factory now takes the budget and renderChunk passes it with the chunk worker count.
* test(producer): run the chunk injector budget test under bun and format
* fix(producer): divide the frame cache budget across workers in the producer, not the engine
The engine package was 13 bytes over its packed-size budget with the worker count threaded through it. The orchestrator and the chunk path now hand the engine each worker's share of the budget, one helper owns the division, and the engine is unchanged.
* fix(engine): encode with the same x264 and x265 thread count on any host size
libx264 and libx265 pick their thread count from the host core count when
ffmpeg is not told one, and the bitstream depends on it, so the same frames
encoded to different bytes on 8, 16 and 32 vCPU nodes and chunks of one
render could differ. Both encoders now pass the thread count from the CPU
quota (Node reads the container limit), which is fixed per pod.
* chore(engine): raise the packed size budget for the encoder thread module
* fix(producer): a blank media src is no source instead of the page itself
A video with src="" reads back from the browser as the composition page URL. The probe then treated index.html as its media file and failed the whole render with 'media sources are text documents'. A blank own src with no <source> child now reports no source.
* style(producer): format the blank-src test
* fix(producer): long renders with small frames no longer fail the up-front disk estimate
The pre-capture check priced PNG frames as raw RGBA, so a long, mostly transparent overlay was refused for needing hundreds of GB it would never write. The measured projection after the first frames already stops a render that cannot fit, using real frame sizes, so the estimate no longer throws.
* fix(producer): drop the comment the ratchet counts
* fix(producer): trim comments so the file's comment share does not rise
* fix(engine): render into a folder whose name holds a percent sign
ffmpeg reads an image sequence path as a printf template, so a folder like
"12% Up" was parsed as a format code and every video render failed. The
folder part of each frame-sequence path now goes through one helper that
doubles the percent sign.
* fix(engine): drop the helper doc line the comment ratchet rejects
* fix(core): a scene paints no filter output from an HTML layer before its start or after its end
* fix(core): a scene paints no filter output from its pseudo-element layers outside its window
* fix(aws-lambda): make webm and mov audio in the separate audio task like mp4
A WebM or MOV plan mixed and encoded its own audio, so on a stack whose plan runs on an arm64 function the audio came out of a different ffmpeg build than an x86 render's. The plan now leaves that audio to the separate audio task, as mp4 already does, and the assembler joins the chunks with that file. Only mp4 chunks stitch; cfr and png-sequence keep their audio in the plan.
* test(aws-lambda): join mov chunks with the audio task's track too
* refactor(producer): keep the separate-audio gate's one-line doc comment
* fix(audio): fit PCM to video frames before AAC delivery
* fix(audio): migrate the CLI diagnostic stage fixture
* fix(test): measure the whole unscaled audio residual
* test(audio): refresh the speed-ramp delivery golden
Tooling and logs need more than "off". Give the runtime's report a reason
code for every way blur can be off (renderer-off, unsupported-browser, each
html-in-canvas defect, capture-failures), and two calls for tooling:
probe() runs the same scan a render would, without painting, and suspend()
stops all painting for the rest of a session without changing the reported
decision.
The engine prints one line per session from what the page reports:
`[motion-blur] on reasonCode=null movers=N failedFrames=K`. The first-frame
line cannot know about later failures, so the session end prints the same
line with "(session end)" when any frame fell back to plain DOM; a clean
session stays quiet. A render where every capture failed no longer reads as
"on" with nothing else.
Add the first test that reads blurred pixels: real Chrome with
html-in-canvas, the engine's frame order (seek, wait, composite, screenshot).
It checks the trail along the motion line, that what does not move matches
the sharp frame, that a box centered with translate is blurred where it
really is, and that a flipped backface card renders the plain frame. It
skips when Chrome or html-in-canvas is missing.
Folded review fixes: R13 (runtime half), R17 (engine half), T1 (the guards
that need reason codes), T3 (log-line and pixel tests).
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Decide the html-in-canvas defects once for the whole timeline instead of at
the frame being drawn. The mover scan already seeks the timeline, so the
defect scan samples along it and a defect that only appears mid-timeline (a
rotated root, a flipped card) turns blur off from frame 0. Defects now block
on every Chrome version, so a render matches the preview it came from, and
GSAP's identity transform on the root no longer counts as one.
A per-frame check stays as a backstop for what scripts add later, and it
costs only that frame: a defect the scan missed (a short root shake, a
backface flip between samples) renders that frame sharp and logs once, and
blur resumes when the defect is gone. A block raised by the scan is lifted
when its cause is removed or the timeline changes; a capture failure stays
for the life of the instance.
The candidates a scan checks include faces whose backface-visibility: hidden
arrives through an animation. Each rescan adds to them and the per-frame
backstop reads the same set, so an animated-only flip card is caught by both.
Look for WebGL canvases in the whole document: a render's shader-transition
compositor appends its canvas to body, while preview puts it inside the root.
Turn page blur off on routes that cannot host it. The HDR and layered
compositors screenshot each DOM layer after hiding the others, which does
not reach blur's canvases, and a distributed (cloud or Lambda) chunk render
has no Studio preview to match and no guarantee of html-in-canvas in the
worker's Chrome. The producer sets pageMotionBlur to hdr-render,
layered-composite or distributed-render there, the distributed plan's
compile-time hint no longer forces screenshot capture, and the runtime says
which route turned it off.
Find the same movers whatever time a scan starts: the baseline is read at
t=0, so each worker of a parallel render picks the same set. An ancestor's
pan is not counted as its child's move.
Folded review fixes: R5, R10, R15, R19.
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Carry the runtime's blur through the render pipeline and the Studio render
panel.
- engine: installPageMotionBlur sets the page-side off switch and the render's
device scale before any author script runs; CaptureOptions.pageMotionBlur
lets a caller stand the page blur down.
- producer: a composition that asks for data-motion-blur on the root the
runtime would pick forces screenshot capture, because html-in-canvas needs
a real paint; the probe session carries the request's pageMotionBlur like
every other session.
- studio-server and Studio: the render panel sends previewMotionBlur, the
render route turns "off" into pageMotionBlur once, and the Studio dev server
and the CLI server both receive the final value.
- mover scan: a large element that is see-through while it moves (a sweep, a
light leak) is no mover, including a large image or canvas with alpha; the
scan skips elements moved by a 3D individual transform property.
- cli snapshot injects video frames before the seek settles, matching the
engine's order.
Folded review fixes: R1, R2, R3 (scan half), R7, R16, R18, T3 (render
adapters).
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* refactor(skills): fold six domain skills into hyperframes and media-use guides
hyperframes-core, -animation, -keyframes, -creative and -studio move to
skills/hyperframes/{core,animation,keyframes,creative,studio}/, and
hyperframes-audio moves to skills/media-use/mix/. Each moved SKILL.md becomes
a frontmatter-free GUIDE.md so installers that scan recursively for SKILL.md
no longer see them as skills; the published list goes from 21 to 15.
Relative links, code paths and skill-name mentions are rewritten to the new
locations, the catalog surfaces, marketplace core-skills list, fallback core
set, module-copy targets and skills-manifest.json follow, and
`hyperframes skills update` now removes the retired skills (RETIRED_SKILLS)
when they are still installed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): slim the core entries and give media-use a HeyGen and a free path
The four entries every run opens are rewritten to what an agent cannot infer:
hyperframes 21.8 KB to 7.7 KB, hyperframes-cli 24.8 KB to 6.9 KB,
hyperframes-registry 9.9 KB to 4.3 KB, media-use 9.7 KB to 4.9 KB. Detail moved
whole into references (routing close calls and creator edits, CLI pin upgrades,
catalog search tiers, CLI conventions and feedback reports).
- hyperframes: state table, routing, the render contract, a terminal work loop
(check, snapshot, preview before render), the desktop-app hand-off, and a
table of the folded guides.
- media-use: one table of HeyGen versus free local tools per need; HeyGen
sign-in, credits and the CLI's video-agent, avatar, clone, lipsync, translate
and clipping commands; every free path works with no account.
- scripts/check-core-skills.mjs holds entries and descriptions to budget and
checks every taught `npx hyperframes` command and path; runs in lint.
- Core entries are exempt from the formatter: its table padding cost ~1.5 KB
per entry on every run.
- Content tests follow the text to its new home.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): plugin runs skip skills update; media-use explains HeyGen sign-in and TTS setup
The smoke eval showed a presenter request never reaching media-use and the
free voiceover failing on a missing kokoro-onnx; plugin runs also ran the
skills update the plugin rules forbid.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(skills): drop a committed Python bytecode file and ignore bytecode
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): align the entry with the Studio skills and keep the workflows' measured facts
The entry drops routing and the intent interview and carries the contract, the
editable-project rules and the work loop. Studio's format, runtimes, motion,
look and captions references come over for terminal use, frame presets move to
assets/, media-use gains Studio's mixing and local-command references, and
film.md and overlays.md keep what the workflow skills measured.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills)!: retire the workflow skills and folded guides; publish seven knowledge skills
The skill pack now publishes exactly seven skills - hyperframes,
hyperframes-cli, hyperframes-registry, media-use, figma, slideshow and
remotion-to-hyperframes - carrying knowledge, not process. The eight
workflow skills, the folded domain guides and the routing/interview
references are deleted; their measured facts already live in the
hyperframes references (format, runtimes, motion, look, captions, film,
overlays) and media-use's mixing reference.
- bake-basemap.mjs and locate.mjs move to skills/hyperframes/scripts/
(bake-basemap gains --help); extract-audio-data.py moves to
media-use/audio/scripts/ for the caption motion reference.
- skills update: the eight workflows join RETIRED_SKILLS so existing
installs prune them (by lock attribution only, since their names are
generic); naming a retired skill reports where its knowledge went and
succeeds instead of failing as unknown.
- Capture prompts point at the hyperframes skill's references/film.md;
the live-map warning names the new bake-basemap path.
- Content tests that pinned knowledge are retargeted to the kept
references; tests of deleted workflows, scripts and process go.
- Catalog surfaces, templates and docs describe the seven skills.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): resolve the local.md captions link and allow capture's output files in citations
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): regenerate the skills manifest from tracked files only
The committed media-use hash included an ignored __pycache__ bytecode file present
in the local checkout, so CI's clean checkout hashed differently.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(docs): restore generated catalog pages and docs navigation to main
Catalog PRs commit item sources only; CI regenerates the pages from the edited
registry sources. docs/docs.json is catalog navigation, so its Workflows group
rename is left for a non-catalog PR.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* build(cli): raise the CLI size budget for the skill references it bundles
The CLI bundles skills/hyperframes, which now carries the folded domain guides and
frame presets (155 KB to 1.08 MB in git). Budgets are the CI-measured size plus 5%:
packed 7526634, unpacked 22593213.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test: keep repo-sweep guards meaningful after the workflow skills left
The last shell ffprobe caller and most root index.html examples lived in retired
skills. The shell detector is now pinned on a literal command line, and the
structure-rule sweep needs more than 5 roots (11 remain) rather than 30.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): slim figma, slideshow and remotion to the Studio shape; restore the storyboard board
slideshow 33.5 KB to 4.9, figma 18.2 to 6.2, remotion-to-hyperframes 11.0 to
4.9, each keeping the facts only it holds (the figma motion helpers, verify gate,
storyboard decoding and usage beacons; present and hand-off; the Remotion lint,
SSIM check and API map). The board page format from the storyboard recipe comes
back as hyperframes/references/storyboard.md: the desktop app's Storyboard tab
reads boards/storyboard.html by its frame, title and time classes. The entry and
the init template now say to snapshot the moments an edit changed before
replying, which the terminal eval found skipped in 22 of 38 runs.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): manifest without local bytecode; Studio's edit prompt names kept files
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): render when the person asks for the file or for no stops, not only after approval
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): keep the voiceover carve in media-use; presets stage their own fonts
Review follow-ups on #5651:
- carve.mjs and its tests move to skills/media-use/audio/scripts; the audio
docs, mixing.md and two Studio comments point there.
- code-editorial's FRAME.md says to copy its bundled WOFF2s and add the
@font-face blocks, instead of promising the retired build-frame.mjs.
- media-use states the host rule before the HeyGen sign-in steps.
- storyboard-format marks transition_in as a note the agent builds itself.
- Two stale citations (creator-editing-recipes.md, sub-compositions.md).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(skills): regenerate the manifest without local bytecode
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): the CLI render step follows step 4's rule, not approval alone
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(ci): run the core-skills check from lint again after the main merge
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): films that tell a story start from three storylines; maps use real imagery
Opus 5.5 arm D read motion.md and applied it, but its launch film was a
feature list (pain, three feature cards, proof, logo): film.md left story
to the model. And the map case never opened film.md, so it drew Natural
Earth outlines instead of baking satellite tiles.
- references/story.md: the generic half of Studio's moment-film story and
craft (one thing, three storylines with world and device, the hook,
beats, worlds, copy, brand, tells of a generated film). The person picks
a storyline, or the agent takes its recommendation when asked for the
file or for no stops.
- The entry's work loop names the storyline step for films over ~15 s; the
references table routes stories to story.md and maps to film.md.
- film.md points at story.md and gives maps and real places a section.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): story.md follows the entry: no voiceover unless asked
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): bake-basemap starts from a globe and stops waiting out its idle timeout
The Opus 5.5 map run spent 13 of its 38 minutes on parallel bakes that
stalled and 15 more recording 300 frames, after rewriting the script because
it could only start on a flat map.
- PROJECTION=globe with a seeded starfield and atmosphere, and FROM/PAN to
turn the globe to CENTER before the dive.
- Each frame polls for loaded tiles instead of waiting for an "idle" event
that had often already fired: 30 frames went from 198-219 s to 10-13 s
locally with byte-identical frames.
- Progress prints seconds a frame and minutes left; film.md shows the globe
call and says to run one bake at a time.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): render in the foreground; wait on a detached render's log, never pgrep -f
Both Opus 5.5 explainer runs started a detached render, then waited with
`until ! pgrep -f "hyperframes render"`, which matches its own loop: each
render took 88-117 s and the wait sat for the tool's full 10 minutes.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): route each kind of film; storylines carry look and sound; add review and sound
The entry routes by request (launch, explainer, music, data, vertical, 3D,
character) to a self-contained file under references/types/, and a film
over ~15 s now gets music and sounds on its visible actions unless the
person asked for silence. Each storyline names its look and its sound, so
picking one picks both; STORYBOARD.md frames carry `reads` and `sound`.
review.md is the whole-film pass: the check warnings, the review sheets,
seven questions to look for, P0/P1/P2 notes and REVIEW.md. motion.md gains
eases by role, six title entrances, a reading-time rule and nine measured
transitions; look.md a design plan before the first scene, with presets
kept for decks and named picks. story.md adds the second default that
generated design falls into, with what to do instead.
The blank template and format.md no longer teach the fade-and-slide-up
entrance on a dark canvas.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(cli): check warns on unread lines, dead stretches, small type, empty frame 0 and silence
A coding agent cannot watch its film, so check now measures what a viewer
notices over time. A new `craft` section holds warnings only (exit 0
unless --strict), sampled on the dense 8 fps grid the overlap pass
already seeks, skipped for --range checks:
- line_unread: a reading-size line (>= 30 px tall at 1080) on screen
under 0.4 s for 1-2 words, else under 3/4 of 0.5 s + 0.3 s a word;
the film's last line needs 2 s. Short lines are re-measured between
samples to ~0.01 s before judging. Captions-track and
data-layout-ignore text are exempt.
- dead_stretch: more than 3 s where the keepsMoving liveness signature
does not change, outside the final 3 s and any data-hold element;
quiet when sweep_static fires or a whole-composition keepsMoving
assertion owns the limit.
- text_below_floor: text under 2.6% of the short side (28 px on 1080)
held 0.5 s, one finding per file listing every element.
- first_frame_empty: frame 0 paints only the background.
- film_silent: over 15 s with no unmuted audio or audible video.
Each message names the fix and the skill reference to read. keepsMoving
now also leaves the film's last 3 s (a still end card) out of its
static runs, so the two agree.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(cli): snapshot --review captures a beat sheet, seam strips and a phone sheet
snapshot --review picks the frames that show a film over time instead of
evenly spaced ones, in the same single browser session:
- review-beats.jpg: one settled frame per scene at its start + 0.3 s,
plus one per timed data-beat="name" element, labelled time and id.
Scenes are the composition's top-level timed clips and sub-composition
hosts, read like `timeline` reads rows (captions and audio excluded)
and timed by the runtime's start resolver over the live page.
- review-seam-NN.jpg: frames at -0.2, -0.1, 0, +0.1, +0.2 s around
every cut (each scene start, and any scene end no scene starts at).
- review-phone.jpg: the beat frames at 360 px wide.
--composition scopes the review to one sub-composition; with fewer than
two scenes the beats spread evenly and a note says so. snapshot also
gains --json: one envelope with every frame's path and time and, under
--review, the scenes, beats and seams behind the sheets; progress moves
to stderr.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(cli): craft warnings stay out of --strict; a hold gets a real next action, not a caret
Craft findings (unread lines, dead stretches, small type, an empty first
frame, silence) are judgment calls a film may break on purpose, so
--strict keeps gating defects only. The check reference no longer
suggests a blinking caret to keep a still frame moving.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(cli): review spreads beats when every scene starts together; map credits skip the type floor
A map film whose basemap and overlay both start at 0 s got one beat
frame; the even spread now keys on distinct scene starts. film.md tells
agents to mark the imagery credit data-layout-ignore and back it, since
the new type floor flagged it on the first real run.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(cli): review frames each scene at its middle too, not only 0.3 s in
On the first real explainer, 0.3 s into a scene showed a bare colour
field before its title entered. Each scene longer than about a second
and a half now adds a frame at its middle, so the sheet shows what the
scene builds.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(skills): a film is built from a numeric SPEC.md, with defaults banned by name and a look test
Studied in full: flxrnc's Motion Lab (vague vs directed prompt), 0xMovez's
course, EianLu's riso skill, agentchengfeng's style samples and the two
Opus 5.5 guides. Their common point: adjectives are where the default look
lives, so the agent now writes SPEC.md before any scene (GOAL, OUTPUT,
TIME, LOOK, MOTION, SHOTS, BANNED, QA), with a number for every adjective.
spec.md bans the defaults by name, unless the brand or reference uses
them (a dark void with small things floating, glow and particles, the
cream-serif-mono-label look, fade-and-slide-up, corner brackets), and
grows that list after the first render. Each storyline gets a 3 s test of
its hero shot before the pick; a look is written as the rules of one
medium. Composition numbers: the hero fills 40% of the frame, one hero
move per shot, the shot's information in within 2.5 s then a held poster.
motion.md adds springs by damping ratio, shorter exits, overlap, stagger
and the beat grid; review.md reviews as a harsh director and grows BANNED;
vertical.md takes the riso skill's measured safe band; explainer.md the
shapes of information.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): story before look; every object reads as itself; posters only for info cards
The first spec-driven C02 swapped the dark default for a flat Swiss
poster, and the story went with it: two black bars nobody read as mirrors,
slogans instead of cause and effect, a storyline picked for its picture.
The spec now serves the story: a storyline is picked for the story a
stranger would follow, the look test asks first whether a stranger
understands what they see, and SPEC.md gains an OBJECTS line giving each
thing the material and details that make it read without a label. The
2.5 s held-poster rule is limited to information cards; in a story film
every shot causes the next. Explainers start from a person or situation
the viewer knows. Review names every object as a stranger would and
treats one that needs explaining as a P0.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(skills): storylines are three films in words, each with a device and a form
The three storylines are pitched as text and the person picks from them;
nothing is built for a storyline before the pick, so the 3 s hero-shot
look test is gone. Ideas pair a device with a form, five or more first,
keeping the three that differ most, and the one stop states the rest in
an assumption line that names every paid item beside its free option.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* ci: list main's new embedded-captions hf-cli test among the retired workflow files
The merge with main removed it with the rest of the retired skill, and the
deletion guard rejects any removal it has not been told about.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(core): keep the audio preset comment within core's size budget
The comment's new citation to media-use's mixing reference made the published
package 14 bytes larger than its budget, which sits at main's exact size.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(cli): craft keeps its own 8 fps grid; comments back within the checks
Review on #5651 found that the merge moved craft onto the overlap pass's
times, which main caps at 180, so a brief line in a long film fell between
samples and line_unread missed it. Craft now builds its own grid at 8 fps up to
600 samples, as before the merge, and a 60 s film with a line shown for 0.2 s
pins it.
The Comments check passes again: REVIEW.md and STORYLINES.md join the files a
skill writes into a project, the comment lines this branch added come back
under each file's share at the merge-base, and the audio preset comment drops
a word the check reads as history.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(engine): a muted video no longer plays at full volume in renders that skip the browser probe
A <video data-has-audio="true" muted> with an audio stream is silent in
preview and in renders that run the browser probe, but a render that
skipped the probe mixed it at volume 1.0, ignoring data-volume. The audio
parser drops muted videos; the video parser kept hasAudio from
data-has-audio alone, so the auto-detect fallback re-added the sound.
The video parser now asks the same isAudibleVideoElement check.
* style(producer): format the muted-video parity test
The AAC true-peak correction assumed each dB of attenuation lowers the
encoded peak by a dB, aiming 0.5 dB under the -1 dBTP ceiling every pass.
FFmpeg 8.1's AAC encoder can hold the peak flat over a few dB of
attenuation (a production render measured -0.5, -0.6, -0.6 dBFS at -3.4,
-4.4, -5.3 dB), so three passes ran out and the render failed. Each pass
now aims twice as far below the ceiling (0.5, 1, 2 dB). The first pass is
unchanged; exhausting three passes still fails the render.
* perf(producer): cloud render chunks start their capture browsers while video frames extract
* refactor(producer): say the chunk extraction overlap in names, not comments
* feat(cli): cloud render uploads heavy media as HeyGen assets
Media of 5 MB or more leaves the project zip: it uploads through the
direct-upload flow as a HeyGen asset and the zip's HTML and CSS point at
the asset URL. Uploads are cached by content hash and re-checked before
reuse. Media the asset API refuses or caps is re-encoded at visually
lossless quality first. The zip must end under 200 MB or the command
stops and names the largest files left.
* feat(cli): cloud render offloads GIFs, still images and script-referenced media
Animated GIFs in the entry composition go through the renderer's own GIF
preparation and upload as the resulting video. Still AVIF, GIF, BMP and
TIFF images convert to PNG. Plain string literals in scripts and JSON that
name offloaded media point at the asset URL. ffprobe calls end their
options with "--", and two comment lines are trimmed for the ratchet.
* feat(cli): cloud render uploads only the parts of long media a composition plays
A video or audio file that has to be re-encoded is cut to the union of the
windows its clips play, with half a second of margin on the source's frame
grid, and the staged composition's media starts shift to match. Clips in
sub-compositions, looping clips, speed ramps and files a script names keep
the full file.
* fix(cli): cloud render no longer decodes media files as text when rewriting references
* fix(cli): cloud render keeps media chosen at render time and encodes the staged bytes
Only media a reference points at is offloaded, so a file picked through
--variables or built by a script stays in the zip. Variable-bound clips and
clips with a media fragment keep their full length. Entry GIFs resolve the way
the renderer rebases a nested entry and are prepared only from 5 MB, and every
encode reads the staged bytes its cache key hashes. Query strings and media
fragments survive the rewrite, a null cache file reads as empty, and the docs
say asset URLs are public and unsigned.
* fix(cli): cloud render comments cite only what this repo holds
* fix(cli): cloud render handles media over 2 GiB and explains a too-large upload
Media of 5 MB or more stays on disk until it is used: it is hashed and uploaded
as a stream, and a file edited after staging stops the run instead of caching
the wrong upload. Files left in the zip are sized before any is read, so a
4 GB file reports "too large" instead of crashing.
The asset API rejects an over-limit upload as invalid_parameter with "File too
large"; that now gets the dry-run hint instead of the --help one.
* fix(cli): cloud render checks and reads each staged file through one handle
Staging, the on-disk staged read and the re-encoded output read each opened the
path twice: once to stat, once to read. A file swapped or edited between the two
could be read without the size or change check seeing it. Each now opens once,
checks the open file's stat before and after the read, and reads from it.
* test(cli): a heavy file a script builds from split strings stays in the cloud render zip
* fix(cli): cloud render keeps media scripts may reach and never trims clips scripts retime
- An offloaded file keeps its zip copy when script code holds both its stem and
extension, so a path joined at runtime still loads.
- Entry GIFs are prepared from a private copy checked against the staged file,
so the GIF cache never stores a later edit.
- No cut when any script can write a clip's timing attributes, dataset,
currentTime, playbackRate, src or loop.
* perf(producer): multi-worker mp4 renders start encoding while frames are still being captured
On the multi-worker disk capture path the encoder used to start only after
every worker had finished. It now starts with capture and is fed frame 0, 1,
2... in order as each frame file lands, with the same encoder arguments, so
the output is byte-identical and the encode left after capture is shorter.
Off for drawElement capture (a failed self-check discards every frame) and
for png/gif/webm/mov/HDR/GPU/chunked/HLS output. If the overlapped encode
fails, the frames are encoded after capture exactly as before.
* perf(producer): encode during capture only when a core is spare for the encoder
With every core busy capturing, the encoder only slowed capture down, so the
overlap now runs only when the render uses fewer workers than the CPU count
worker sizing sees. Also pins the reviewed producer test-lane source hash and
drops an added comment line.
* refactor(producer): shorter comments on the encode-while-capturing helpers
* fix(core): grade an image that appears after its scene starts hidden
A graded <img> was only picked up if it was visible when grading scanned the page. An image in a
layer revealed later by a timeline, or in a sub-composition whose host starts later, stayed
ungraded in snapshot and render. Each draw pass now adopts graded media that has come into view.
* fix(core): keep the grading file's comment share flat
* fix(core): keep a grade switched off live from coming back on the next frame
Miguel Ángel