* fix(cli): the browser the CLI launches exits with it, even on kill -9
* fix(build): the root build runs on Windows
* perf(cli): a repeat check runs in a process that already has the CLI and Chrome up
* fix(cli): a warm check delivers all its output before it exits
* fix(cli): the warm check only talks to a server that proves it is the user's
The warm socket now lives in a private, owner-verified directory (XDG_RUNTIME_DIR,
~/.cache or LOCALAPPDATA) next to a 0600 token. Client and server prove the token to
each other over fresh nonces before any request is sent, the request carries an
HMAC of the environment instead of the environment itself, and frames are capped.
* fix(cli): a warm check prints its stdout even after something went to stderr
The client stopped writing frames once any frame had been written, so a check that
printed a notice to stderr first lost its JSON report.
* test(cli): list the warm-check client fixture as a fallow entry instead of an inline ignore
* fix(cli): a check started while a warm one is busy runs cold at once on Windows
The warm server stopped listening when it accepted a client. On Windows the pipe name stays alive through the served connection, so a second client's connect saw a busy pipe and waited 30 s for a free instance before running cold. It now keeps listening while serving (accept already declines anyone else) and stops when the served check finishes.
* test(cli): the orphan test reads the check's own Chrome and watchdog instead of racing a poll
On Windows each process listing is a PowerShell call of about a second, so a check that finishes quickly could start and close Chrome between two polls, and the test failed its own 'Chrome was seen' precondition. A fixture now records the Chrome tree and the watchdog from inside the check when its first page opens.
The root (and propagated body) background paints the page canvas whatever its
visibility, so an animated page background could make a still scope read as
moving. Scoped captures now clear those backgrounds and keep the resulting
transparency, and the frame comparison reads alpha, so dark paint revealed on
the cleared page still counts.
* fix(cli): check no longer calls films moved by SVG masks, strokes or canvas static
The frozen-sweep guard compared DOM-channel fingerprints only, which never see
SVG mask and clip contents, stroke dashes, or canvas and WebGL pixels. Before
reporting sweep_static, check now compares the rendered frames at the same
sample times; a still verdict stands only when every frame is identical.
* fix(cli): keepsMoving also trusts rendered frames, compared above render noise
keepsMoving read the same DOM liveness channels as the sweep guard, so mask,
stroke and canvas motion failed it as frozen. A span it calls frozen now gets
a frame capture every half maxStaticSec, and spans split where the frames
change. Frames are compared in the page, decoded like the contrast pass, and
count as moved only when at least 16 pixels change by more than 24/255, so
antialiasing and dither noise cannot make a still film pass.
* fix(cli): scoped frame checks see only the scope's own paint
A keepsMoving scope's capture now hides every box outside the scope's subtree,
so a still title over a moving background stays frozen. A span marks only the
frames that still hold its signature, so a span that starts at a cut keeps its
first frame. A scope that matched nothing stays motion_selector_missing. The
sweep counts a frame change only if the first sample time still renders its
first frame, so wall-clock paint is not read as motion under seek.
Color grading paints a video or image into a canvas the runtime inserts right
after it. The layout audit treated that canvas as authored canvas art: the
video's data-layout-* opt-outs never reached it, and canvas_content_at_edge
read full-bleed footage as drawn content cut by the frame. Opt-out lookups now
resolve a grading canvas to its source element, and the edge rule leaves
grading canvases alone, as it does the ungraded video.
* fix(cli): check, snapshot and Studio read project text as UTF-8
The preview servers sent HTML, CSS, JS and JSON with no charset, so Chrome
read a page without <meta charset> as windows-1252 and stills showed garbled
non-Latin text while the render, whose file server already sends utf-8,
was correct. The shared MIME table now carries charset=utf-8 for text types,
and the two hard-coded headers use it.
* fix(cli): snapshot shows accented text when the page's meta charset sits outside its head
A <meta charset> past Chrome's first 1024 bytes and outside <head> is ignored, so with no
charset in the header the page decoded as windows-1252. The browser test serves such a page
through the static server snapshot uses. play compares the media type without its charset,
so JSON assets keep their content-hash revalidation.
* perf(cli): commands run without esbuild, puppeteer-core or sharp until they need them
A host that bundles the CLI can now leave these packages out: each command
loads them only where it uses them, and sharp installs once on first use when
it is missing. The build check fails on any static import of them again.
* fix(cli): use a host's sharp from NODE_PATH and let hosts turn first-use installs off
* fix(cli): capture fails when sharp is missing, and one install serves every caller
A missing sharp now stops the asset download up front instead of counting every
image as unavailable. Concurrent first uses share one install, a failed install
is not retried in the same run, and preview thumbnails report why. Hosts can turn
off installs for a list of packages, e.g. HYPERFRAMES_NO_OPTIONAL_INSTALL=sharp.
* fix(cli): snapshot --describe still works without sharp, resizing only when it is installed
* style(cli): format the sharp loader and its tests
* build(cli): sharp is an optional dependency, so a host can ship the CLI without it
* build(studio-server): keep sharp out of the bundle now that it is optional
* perf(cli): check splits its samples over a second Chrome when a core and memory are spare
* fix(cli): check keeps its result when the second Chrome fails, and counts container memory
* fix(cli): check's helper Chrome code keeps the files' comment share
* fix(cli): check's sampling comment fits the line limit
* perf(cli): snapshot decodes phone video frames at their on-screen size, several at once
* perf(cli): snapshot decodes phone video frames at their on-screen size, several at once
* test(cli): type the displayed-size snapshot cases instead of splitting strings
* feat(cli): lambda deploy works from the npm install, no repo checkout
Each release builds the Lambda handler zip on Linux, attaches it to the
GitHub release, and pins its URL and sha256 (plus the SAM template) in
@hyperframes/aws-lambda before npm publish. Outside a checkout, lambda
deploy downloads that zip, refuses it on a sha256 mismatch, and deploys
it with the packaged template. With neither a checkout nor a pinned zip
it fails up front with one message naming every way out.
* fix(cli): lambda deploy test mocks the pinned-handler reader, comments fit the checks
* test(cli): sam tests deploy from a SAM directory, not the repo root
* chore(aws-lambda): raise the package size budget for the pinned handler template
Measured with pnpm@10.17.1 pack on Linux. Release state (prebuilt/ written by pin-handler-zip): 77,307 packed / 285,654 unpacked. CI state (no prebuilt/): 70,104 / 253,390. Packed is the check's 5% headroom over the release size. Unpacked is 292,000 (2.2% over release) because a 5% budget would put the CI pack under the 85% ratchet.
* test(cli): anchor the docs link check in the lambda deploy fallback test
* fix(cli): lambda deploy logs the verified handler sha256 and keeps a presigned query out of errors
* fix(cli): a redirect-limit download error keeps a signed query out of its message
* test(cli): the redirect-limit leak test types its rejection
* docs(aws-lambda): header comments cut to what the code cannot say
* chore(aws-lambda): size budget fits both the CI and the release package
* fix(aws-lambda): old stacks are refused before any upload; stale local copies never re-upload
* docs(aws-lambda): where the CLI keeps a deployed stack's state
* perf(producer): distributed chunks download only the videos they show
A project video under compiled/ that chunks extract their own frames from
now goes only to the chunks that show it, like video-sources/ files. A
chunk skips the readiness wait for a deferred video it does not show, so
it no longer logs media_load_failed for a file it was never given.
* fix(producer): cloud site uploads pack the files their id hashes, symlinks included
* perf(aws-lambda): chunks read only the byte ranges of a video they need
A chunk that extracts its own frames from a deferred project video now
reads it from S3 by byte range through a short-lived pre-signed URL,
instead of downloading and hashing the whole object first. S3 attests the
object's stored SHA-256 before the URL is signed; every other artifact is
still downloaded and verified as before.
* perf(aws-lambda): upload project files once to a shared content-addressed area
* chore(producer): keep planV2 comment share at the base level
* fix(engine): verify the TLS certificate of https video inputs and never surface their signed query
* fix(aws-lambda): renders re-upload a site's expired files instead of failing
* perf(aws-lambda): the plan step probes project videos in place instead of downloading them
* fix(engine): https video reads reconnect and time out, and redaction covers wrapped causes
Signed-URL reads now pass -rw_timeout, -reconnect and -reconnect_delay_max (both Lambda builds accept them) and ffmpeg also -reconnect_on_network_error, so a transient S3 error or a stalled read fails fast or recovers instead of waiting out the process timeout. Query redaction now descends into error.cause.
* fix(aws-lambda): share entry bundles and ratchet package budget
* fix(aws-lambda): site ids uploaded before files.json still render
A site id from an earlier sites create has only sites/<id>/project.tar.gz.
renderToLambda now falls back to that tarball, which the plan step still
reads, and when neither key exists it fails before starting with both keys
named and the fix (run sites create again).
* fix(aws-lambda): renders fail before upload on a stack too old to read sites
A stack deployed before files.json sites untars the site file in its plan
step and fails after four retries. renderToLambda now reads the state
machine definition first and stops with STACK_TOO_OLD and the redeploy
command when the plan task does not list SITE_MANIFEST_INVALID, which new
SAM and CDK stacks now treat as terminal. The README's upgrade section
drops the planProtocol v1 fallback, which hits the same failure.
* docs(aws-lambda): document shared content storage and the upgrade path
The deploy guide still described the tarball HeadObject short-circuit. It
now covers the cas/sha256/ store and its 30-day lifecycle rule, the
STACK_TOO_OLD, SITE_FILES_MISSING and SITE_NOT_FOUND errors, that old site
ids keep working, and that every project uploads once more after the
upgrade. A test keeps a troubleshooting entry for each pre-submit error.
* fix(producer): chunks log each video they read by URL, without its query
A chunk reading a video straight from the bucket left no trace in its log.
It now logs one line per remote source with the path and the URL minus its
query string, so signatures never reach the log.
* fix(engine): ffmpeg errors never keep part of a signed URL's query
A process's stderr tail was trimmed before its URLs were redacted, so a
cut inside a signed URL left a query fragment the URL pattern no longer
matched, and the fragment reached the error. The stderr tail is now
redacted before every trim, and the redactor also strips a signing
parameter cut off from its URL.
* fix(aws-lambda): comments name the site manifest instead of a file path
* fix(engine): stderr redaction survives chunk splits, and every ffmpeg spawn uses it
The stderr tail was redacted on each chunk and then cut, so once a cut
severed a long URL's scheme, the rest of its signature in the next chunk
matched neither pattern. The tail is now kept raw with 2x slack and
redacted only when the text is produced, then cut to size. The unused
producer audioExtractor (its own spawn, unredacted stderr) is deleted and
the parity harness's PCM decode goes through ManagedChildProcess; a test
fails on any ffmpeg/ffprobe spawn in engine, producer or aws-lambda that
does not.
* chore(producer): record the audioExtractor deletion as intended
* fix(engine): drop a signature fragment the stderr tail cut from its URL
The raw tail is kept at twice the cap, but that cut can itself land inside a
signature, and when redaction shrinks the rest below the cap the fragment
survived. After a cut, the severed token is dropped before redaction.
* refactor(engine): name the severed-token drop instead of commenting it
* test(engine): a cut-off URL leaves nothing behind; aws-lambda size budget to measured
* fix(producer): chunk URL-read tests use main's frame list; producer size budget to measured
* chore(producer): size budget at the measured size, no headroom
* chore(producer): size budget back to measured x1.05, the check's own rule
* fix(cli): non-English transcription reaches whisper, and whisper failures say why
transcribe --language <non-English> runs an installed whisper (Parakeet picks its own language); whisper models are pinned by size and sha256 and re-downloaded when damaged; a whisper-cli failure names its build, model file and last stderr lines without paths; the unused --dtw flag is dropped.
* docs(cli): --engine and --language say a non-English language runs an installed whisper
Also strips paths that follow a colon or a bracket from whisper-cli's error tail.
* fix(cli): catalog search never waits on installing the on-device runtime
A plain catalog query on a machine with on-device search set up used to
npm-install onnxruntime-node inline when it was missing, before returning
any results. Only --on-device installs it now; otherwise the search runs
on word matching and says how to turn the on-device tier back on.
* fix(cli): the first-time on-device yes names the command that sets it up
After a yes at the thin-results offer, a plain re-run no longer installs the
runtime, so the follow-up line now names `hyperframes catalog --on-device`
instead of telling the user to run the search again.
* fix(cli): the on-device setup line names the search to repeat with --on-device
* fix(cli): catalog --on-device alone sets up on-device search, so the setup line needs no query
The setup line repeated the user's query, which no quoting keeps intact across
every shell and the argument parser. --on-device without a query now runs the
setup, and both guidance lines name that fixed command. The test runs it through
the real parser.
* perf(cli): check reads a solid text backdrop from styles instead of a screenshot
* perf(cli): check starts Chrome while lint and bundling run
* perf(cli): the CLI reuses compiled bytecode between runs
* test(cli): browser shutdown tests stop their Chrome when a case fails
* fix(cli): snapshot no longer transcodes every iPhone clip it never plays
snapshot served Studio's media proxies by default, so its page server transcoded every browser-hostile clip (every HEVC clip) in full at load. Snapshots show frames extracted with FFmpeg and never use those copies. Proxies are now served only with --proxy; hyperframes.json media.autoProxy no longer turns them on for snapshots.
* docs: snapshot makes no media proxy unless asked
* perf(cli): check and inspect skip paint waits and screenshots nothing reads
* perf(cli): check runs its dense overlap pass in one page task so no frame is painted between seeks
* perf(cli): check waits for late images and fetches instead of a fixed 1.5 s after load
* fix(cli): check gives an event it can only place within the dense pass that pass's time range
* perf(cli): check waits for image decodes before a screenshot instead of a fixed 120 ms
* perf(cli): check screenshots only the text contrast measures and skips samples with none
* perf(cli): seeks check fonts on the settle's second frame instead of waiting a third
* docs(cli): the contrast audit header says how a clipped screenshot is read
* docs(cli): cut the font-wait and contrast-audit comments to short blocks
* fix(cli): timeline delete --start/--end removes only that range
timeline delete accepted --start and --end but never read them, so a range
delete removed the whole clip. It now cuts out only that range, the same way
split, split, delete would, and keeps the parts before and after.
Every edit verb shared one flag set, so a verb silently ignored flags it does
not use. Each verb now declares only the options it reads, from one table that
also checks apply-plan keys, so an unused flag or key is refused before
anything is written.
* refactor(cli): name the per-verb option table instead of commenting it
* perf(aws-lambda): run chunks through a Distributed Map to lift the 40-chunk cap
The inline Map runs at most 40 chunk renders at once. Both state machine
definitions (CDK and SAM) now use a Distributed Map with STANDARD child
executions, capped at the function's reserved concurrency so chunks past it
queue instead of being throttled. Progress reads each chunk's child
execution history so frames, cost and chunk errors still show up.
* fix(aws-lambda): retry Lambda throttles for minutes and stop on oversized chunks
A throttled task used to burn its four catch-all retries in about 30 s.
It now gets its own jittered backoff of several minutes, because another
render on the same function can still hold the reserved slots. Plan
treats CHUNK_EXCEEDS_WORKER_TIMEOUT as terminal in both definitions.
* chore(aws-lambda): trim stack comments to keep the comment share flat
* fix(aws-lambda): keep render --wait working at hundreds of chunks
Progress polls read the chunk Map Run's counts (one DescribeMapRun) instead of
every child execution's history, take exact chunk frames and cost from the Map
output once it exits, and read failed children only when the counts show a
failure. The SDK client retries throttled calls up to 8 times with backoff.
* feat(aws-lambda): size each render's concurrency to the account instead of a fixed 8
* fix(lambda): keep the concurrency SDK external in published modules
* fix(cli): keep an existing stack's reserved concurrency on redeploy
Redeploying without --concurrency sent ReservedConcurrency=-1, which
removed the cap of 8 that earlier CLI versions reserved by default and
stopped recording it in the state file. The override is now left out
when the flag is omitted, so SAM keeps the stack's current value (a new
stack still gets the template's unreserved default). The state file
records the deployed value, and deploy prints the kept reservation.
* fix(aws-lambda): keep progress working for callers without DescribeMapRun
Callers whose IAM policy came from an earlier `lambda policies user`
lack states:DescribeMapRun, so reading a running Distributed Map threw
AccessDeniedException out of `lambda progress` and `render --wait`
while the render itself went on to succeed. getRenderProgress now
catches only that error around the map-run read, returns the parent
snapshot (status, plan and assemble progress) and adds a warning to
re-apply the policy, which both commands print. Other errors still
throw.
* fix(cli): let the documented deploy policy drop a reserved concurrency
When a redeploy removes a function's reserved concurrency, CloudFormation
calls lambda:DeleteFunctionConcurrency. The policy printed by
`lambda policies role|user` granted only PutFunctionConcurrency, so the
update failed and rolled back. Grant it, and pin it with
states:DescribeMapRun in the required-actions test.
* fix(aws-lambda): read smoke chunk hashes from the Distributed Map output
With chunks in a Distributed Map, their task results live in child
executions, so the smoke gate's chunk-hash check found 0 chunks in the
parent history and failed every v1/v2 comparison. Read Chunks[] from the
RenderChunks / RenderChunksV2 MapStateExited output instead, as
getRenderProgress does. The stack this script deploys only has the
Distributed Map, so the inline-Map path is gone.
* fix(producer): cap a pinned chunk size at the worker concurrency, not 16
With --chunk-size or --target-chunk-frames and no --max-parallel-chunks,
the plan still capped the render at 16 chunks however many workers the
account allowed, so --chunk-size 30 on a 60 s render gave 16 chunks of
113 frames. The cap is now one wave across the worker concurrency, as
the --max-parallel-chunks docs say; 16 remains when the concurrency is
unknown.
* feat(cli): remove a stack's reserved concurrency with --concurrency=none
Since an omitted --concurrency now keeps the stack's reservation, there
was no CLI way back to the unreserved default. --concurrency=none sends
ReservedConcurrency=-1, the state file records the stack as unreserved
and deploy says the reservation was removed. Positive N and an omitted
flag behave as before; any other value is rejected instead of being
silently ignored.
* chore(aws-lambda): raise the package size budget for account-sized concurrency
Measured with pnpm@10.17.1 pack after rebasing on main with the S3 assembly change: 67,763 packed / 244,685 unpacked; the budget is the check's own 5% headroom over that.
* fix(core): color grading applies to canvas compositions
data-color-grading on a <canvas> was ignored by the runtime and refused by
lint, the Studio panel and the CLI writer. One source-tag list in the grading
contract now feeds all of them. A canvas is sized from its drawing buffer and
re-uploaded on every seek and transport tick.
* fix(core): keep the comment share of grading files at the base
* perf(cli): cloud render checks render status every 2 seconds instead of 10
A finished render was noticed up to 10 s late. Polling every 2 s cuts that
wait to at most 2 s and stays well under the status endpoint's rate limit.
`--poll-interval` still overrides it.
* perf(cli): cloud render downloads the finished video over parallel connections
The first request asks for the first 8 MB. When the server answers 206 with
the full size, the rest of the file is fetched in up to 7 more ranged
requests at once and each part is written at its offset into the staged
file. A server that ignores Range (200), a small file, or an empty file
is still fetched in one request. A failed part cancels the others and the
existing output is kept, as before.
A background preview detaches, so one started by an agent inside a host app's
run kept running after the run and the app ended, and held the project's
history lock for its whole life. Inside a host app's run, preview now refuses
to start in the background and points to the host app's preview or
--foreground, which stays attached to the run.
* fix(cli): check and snapshot wait for a slow CDN script as long as render does
check, snapshot, layout, motion-shot and validate gave the page 10 s to reach
DOMContentLoaded while render gives it the engine's page-navigation timeout
(60 s), so a parser-blocking script from a slow CDN failed the diagnostics
with "Navigation timeout of 10000 ms exceeded" on projects that render fine.
The separate 10 s budgets are gone: every caller takes the engine's value
(and its env override), raised only by a command's own longer --timeout.
The cold-Chrome allowance for sending the first request is unchanged.
* fix(cli): check and snapshot read only the page-load budget from the engine
resolvePageNavigationTimeout() owns the page-navigation env read for render
and the diagnostics, so check and snapshot no longer resolve the whole render
config, which printed render's Windows streaming-encode warning. validate's
comment and timeout hint now say render shares the budget.
The capability catalog listed correction controls as "adjustments" and
finishing controls as "finishing", while the grading payload calls them
"adjust" and "details", and no capability showed a payload. Agents wrapped a
correction in "adjustments" and the validator rejected it with a hint that
named no key. The schema now owns a correction example; the catalog names its
sections after the payload keys and carries that example, correction and
finishing show an apply payload from it, the unknown-key hint lists the real
top-level keys, and a test validates every grading payload in docs and skills.
* fix(cli): check and snapshot keep their record while another app has the history open
* test(cli): a held history no longer drops the snapshot record
audio and timeline joined a clip's src to its composition's folder by hand,
so a project-root asset used from a nested scene, or a percent-encoded file
name, was "not found" although preview, check and render load it.
resolveAuthoredAssetFile now owns that lookup (the sub-composition rewrite,
then the decoded project-root resolve), and audio, timeline and
media-treatment all use it.
* fix(cli): catalog loads registry items a few at a time so a slow link no longer times them out
* chore(cli): keep the moved concurrency helper's doc to one line
* chore(cli): wrap the concurrency helper's doc under the line limit
* fix(studio-server): send Studio's history edits to the process that owns the project's history
* fix(studio-server): a forwarded history call gives up after 5 s, so a stopped owner cannot hang
lambda reported any ERR_MODULE_NOT_FOUND as '@hyperframes/aws-lambda is not installed', and cloudrun matched its package name anywhere in the message, which also names the adapter when one of its own dependencies is missing. Both now use one helper that matches only Node's "Cannot find package '<pkg>'" and rethrow anything else.
lambda and cloudrun render at 30 fps when data-fps is outside the cloud set (for example 50 or 30000/1001) and --fps is not passed. They now say so and point at --fps. The 30 fps default has one owner, defaultCloudFps.
The integer rates Lambda and Cloud Run accept were written out by hand in
about 25 places across core, engine, producer, the adapters and the CLI.
They now come from DISTRIBUTED_FPS / DistributedFps / isDistributedFps in
@hyperframes/core. Engine config snapshots accept 25 like the rest.
* fix(cli): check's dense overlap pass takes at most 240 samples, at clip windows first
* fix(cli): a held overlap on a long film is still an error under the 240-sample bound
Lambda and Cloud Run renders refused 25 fps although it is an integer
rate like 24, 30 and 60, and PAL footage is 25 fps. The distributed fps
set is now {24, 25, 30, 60} in the producer's config validation, the
plan v2 manifest, the render request mapping, the Lambda and Cloud Run
event types, the CLI's lambda and cloudrun commands and the regression
harnesses.
Co-authored-by: Cursor <cursoragent@cursor.com>
## What a user can do now
Render a video at 720p or 480p: `hyperframes render --resolution 720p` (or `480p`, or the presets `landscape-720p`, `portrait-480p`, `square-720p` and so on) gives a 1280×720 or 854×480 file from a 1080p composition, and the same presets work through the studio-server render route. Before, any preset smaller than the composition was refused ("Downsampling via --resolution is not supported").
## How
- `parsers/src/types.ts`: six presets, `{landscape,portrait,square}-{720p,480p}`, and the `720p` / `480p` aliases (aspect-agnostic, like `1080p` and `4k`). 480p at 16:9 is 854×480: 480 × 16/9 is 853.33, and 4:2:0 encoders need even sides.
- `checkOutputResolutionCompatibility` now allows a preset smaller than the composition. A smaller preset may be off by under one pixel on its long side (the 854); a larger one must still match exactly and be a whole multiple, as before. The aspect remap that `--resolution 1080p` / `4k` apply on their own (`suggestMatchingPreset`, used by the CLI preflight and the compile stage) returns only a preset in the asked tier, never another: with the looser match for smaller presets, a 1922×1080 composition asked for `4k` would otherwise have been remapped to 480p. It is refused instead, as before.
- `resolveDeviceScaleFactor` returns the long side's ratio, so a smaller preset renders at a device scale factor under 1: Chrome lays the page out at composition size and rasterizes each frame straight at 1280×720 or 854×480, the way 4K already renders at 2. For 480p that is 854/1920: the width lands on 854 exactly and the height's 480.375 rounds to 480, so no pad column is added. Output sizes round as Chrome rounds the screenshot.
- The streaming encoder was given the composition's size instead of the output size on both streaming paths (`renderOrchestrator.ts`). 4K never noticed, as both sizes are even; at 480p the encoder skipped its even-size pad and libx264 refused an odd frame (`width not divisible by 2 (853x480)`). It now gets the output size, as the disk-frame encode already did.
- The paths that refused DPR > 1 (BeginFrame, drawElement, the Linux fast-capture gate) now refuse any DPR other than 1, since none of them honor a scale; scaled renders take the screenshot path, as 4K does.
- `render/outputScaleRoute.ts`: the layered compositor (HDR content, and shader transitions in webm and mov) sizes its layers and encoder from the composition and never scales, so a scaled output on that route now stops with a named error, checked where every route and replan is settled (`syncCapturePlan`, beside motion blur's route check). Before, a 1080p request from a 4K composition with shader transitions as mov would have encoded 3840×2160. HDR with `--resolution` was already refused in preflight; 4K on this route had the same gap on main and gets the same error.
## What I measured
- Real renders from source on a Linux box (`bun src/cli.ts render registry/examples/nyt-graph --resolution 720p|480p|1080p`), checked with ffprobe: 720p `h264, 1280x720, yuv420p, 15.0 s`; 480p `h264, 854x480, yuv420p, 15.0 s`; 1080p unchanged at 1920x1080. A frame of each, looked at: same chart and layout, crisp text, and the 480p right-edge column is the page's own color (250,248,242), not a pad.
- Before the encoder-size fix the 480p render failed with `[libx264] width not divisible by 2 (853x480)`; after it, it renders.
- `outputScaleRoute.test.ts`: the real route choice for a 4K composition with shader transitions as mov (`outputSupportsPageSideShaderCompositing`, `shouldUseLayeredComposite`, `createCapturePlan`, `resolveDeviceScaleFactor`) lands on `hdr_layered` at scale 0.5 and is refused. CLI preflight and the compile stage both refuse `4k` on a 1922×1080 composition.
- fails without the fix: without the same-tier rule, the preflight and compile tests for 1922×1080 at `4k` pass the render through at 854×480; without the route check, the mov test does not throw.
- At d33c4458: producer `outputScaleRoute` + `compileStage` 18/18 three runs in a row, `shared` (bun) 9/9, cli `render` 104/104 once and its new test three runs in a row; tsc clean in producer and cli. Earlier at e72f6507: parsers `outputResolutionCompatibility` 20/20, core `index` 26/26, producer `resolveDeviceScaleFactor` 14/14, cli `render` + `parseOutputResolution` 122/122, studio-server `render` 53/53. The new cases fail on main (the presets don't exist; smaller presets are refused). tsc clean in parsers, engine, producer and cli; oxlint and oxfmt clean.
## What I did NOT exercise
- Distributed (Lambda / Cloud Run) renders at 720p/480p: the chunk worker's sizes round the same way, but I did not run a distributed render.
- Studio's render queue keeps its own copy of the rule (`RenderQueue.tsx` `scaleApplies`) and offers no 720p/480p option; unchanged here.
skills update (bare or named) and bare skills now exit 0 with a one-line
notice instead of installing or pruning when HYPERFRAMES_SKIP_SKILLS=1, so a
pinned skills install survives the router's trigger-time update. Closes#5695.
* fix(cli): page loads stop timing out while a busy machine's Chrome warms up
* fix(cli): keep the page-load helper comment-free and its files in the bench filter
* feat(cli): judge text behind a transparent video by its decoded alpha in check
The layout audit treated every <video> as opaque over its whole box, so a
caption between footage and a remove-background cutout always failed
text_occluded and needed data-layout-allow-occlusion, which also hid a word
the subject's head fully covered.
check now draws the video's current (seeked) frame into one reused,
downscaled scratch canvas, once per sample and only for a video a text probe
lands under. A fully opaque frame keeps today's behaviour. A frame with
transparency is judged per rendered word against its alpha: partial cover is
an info finding carrying the most covered word's share, and a word 80% hidden
or two neighbouring words each half hidden is an error. A frame that cannot
be read falls back to opaque with a video_alpha_unread info note.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(cli): serve the video-alpha fixtures from memory, not request paths
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(cli): check finds words under narrow, mirrored or backed cutouts
Review follow-ups on #5657:
- A matte the probe walk passes over, even at clear-alpha points, now triggers
the per-word check, so a feature between two probe columns (an arm, a mic
stand) still fails the word it hides.
- Mirrored videos and images map viewport points through the flip; a rotated,
skewed or 3D-transformed video falls back to opaque with video_alpha_unread.
- A video with an opaque CSS background keeps its whole box occluding.
- cli.mdx lists every unread cause and why a fully transparent frame is one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(cli): render telemetry reports how busy the machine's CPU was
render_complete and render_error now carry system-wide CPU busy share (average
and busiest 2 s window), cores busy in that window, and the one-minute load
average at start and end, for CLI and Studio renders. Nothing is sampled when
telemetry is off.
* test(cli): render tests use a private project folder instead of /tmp/project
* feat(cli): audio answers beats, downbeats, bpm, sections and describe through a provider ladder
Beats and downbeats try api, apple-mu, beat-this, then signal; sections try api, apple-mu,
then signal; describe tries api, then apple-sound. bpm is read off whichever provider placed
the beats. Each provider says whether it can run here and why not, and the ladder warns
when it passes one over.
The signal rung is a TypeScript tracker: log-band spectral flux, autocorrelation tempo with
a log-normal prior, Ellis dynamic-programming beats with a windowed local tempo, a low-band
bar phase, and Foote novelty sections snapped to beats. It reports no steady beat when the
onsets do not repeat.
The api rung posts the audio to HYPERFRAMES_AUDIO_API_URL with HYPERFRAMES_AUDIO_API_KEY and
moves each answered time onto the nearest signal beat within a quarter beat, reporting what
moved and what did not. beat-this runs Beat This! from a configured Python and local weights
and never downloads. apple-sound runs macOS's sound classifier through a Swift helper.
apple-mu reports that Music Understanding needs macOS 27.
words are no longer accepted: they come from hyperframes transcribe.
* docs(cli): audio's beats, downbeats, bpm, sections, describe and their provider ladder
* feat(cli): beats tracks the music with the TypeScript beat tracker, without Chrome
hyperframes beats and Studio's beat detection now share analyzeMusicFromSignal: band energies,
then the spectral-flux tracker in audioRhythm.ts, then each beat's loudness for its strength.
The CLI decodes with the same ffmpeg path as hyperframes audio, so beats no longer launches
headless Chrome. Studio still decodes in the browser and passes every channel.
The old analyzer averaged bpm-detective's tempo with an energy-peak tempo it had only compared
after folding both into one octave, and phased a rigid grid from noisy peaks. On 22 real tracks
with exact script grids it found the tempo 18% of the time with beat F 0.437; this finds it
91% of the time with F 0.887, and refuses the other two rather than guess.
Below a tracker confidence of 0.2 no beats are reported, so ambient or rubato music gets no
invented grid and the CLI still writes nothing. bpm-detective, the energy-peak grid, the
headless analyzer and its prebuilt bundle are gone.
* fix(core): tempo reads 96 BPM off beats on 10 ms frames, not 95.2
The tracker places beats on 10 ms frames, so a 625 ms beat lands 620 or 630 ms apart and the
median interval reported 95.2 or 96.8 BPM. Tempo now averages the intervals within 10% of the
median: as robust to a skipped or doubled beat, without the frame-sized step. Both
hyperframes audio and hyperframes beats read it.
* fix(cli): remove redundant audio context comment
* fix(core): remove redundant tempo interval comment
* fix(ci): record beat analyzer removals and browser evidence
* fix(core): keep flat audio novelty in one section
* chore: keep PR capture evidence out of the source tree
* fix(cli): honor BPM provider pins before analysis
* feat(cli): audio answers beats, downbeats, bpm, sections and describe through a provider ladder
Beats and downbeats try api, apple-mu, beat-this, then signal; sections try api, apple-mu,
then signal; describe tries api, then apple-sound. bpm is read off whichever provider placed
the beats. Each provider says whether it can run here and why not, and the ladder warns
when it passes one over.
The signal rung is a TypeScript tracker: log-band spectral flux, autocorrelation tempo with
a log-normal prior, Ellis dynamic-programming beats with a windowed local tempo, a low-band
bar phase, and Foote novelty sections snapped to beats. It reports no steady beat when the
onsets do not repeat.
The api rung posts the audio to HYPERFRAMES_AUDIO_API_URL with HYPERFRAMES_AUDIO_API_KEY and
moves each answered time onto the nearest signal beat within a quarter beat, reporting what
moved and what did not. beat-this runs Beat This! from a configured Python and local weights
and never downloads. apple-sound runs macOS's sound classifier through a Swift helper.
apple-mu reports that Music Understanding needs macOS 27.
words are no longer accepted: they come from hyperframes transcribe.
* docs(cli): audio's beats, downbeats, bpm, sections, describe and their provider ladder
* fix(cli): remove redundant audio context comment
* fix(core): keep flat audio novelty in one section
* fix(cli): honor BPM provider pins before analysis
* fix(cli): retire old snapshot run folders
* chore: satisfy snapshot comment limit
* fix(history): read and release snapshot leases through verified handles
* test(history): run lease races in Node and check released markers
* fix(studio): verify snapshot leases from open descriptors
* feat(cli): hyperframes audio measures where sounds land, in film time
Adds @hyperframes/core/audio-analysis: an energy envelope, spectral-flux
and energy onsets, the moments a sound lands (onset, steepest rise, peak,
tail end), BS.1770 loudness with true peak, silence spans and a first
guess at sfx/voice/music, all pure functions of PCM. Each capability runs
through a provider ladder so model or service providers can join later.
The radix-2 FFT moves out of audioCarve so both share it.
`hyperframes audio` decodes with FFmpeg and analyzes files, or the film's
own audio and video clips by element, second, range or boundary, reading
only the part of each file that plays. Timeline rows now carry
`mediaStart`, moved by an in-point host's cut, so file time maps to film
time.
* feat(cli): audio placements carry the element's own data-start, local to its composition
* feat(cli): audio placements carry the element's own data-start, local to its composition
* feat(cli): audio places a file with --place-at, drops --boundary
--start becomes --place-at and --media-start becomes --skip-file-seconds,
which only goes with --place-at, refuses to skip the whole file, and warns
when the skipped part held the file's loudest moment. --boundary is gone:
--at and --range cover the sounds a cut touches.
* fix(cli): pass a sound file to ffprobe as a file, on every platform
ffprobe treated a name starting with a dash as an option because the argument
list never ended the flags. The path check compared a Windows drive path to a
POSIX string, and a new timeline comment raised that file's comment share.
* fix(cli): hear a reversed stereo channel, and warn when its peak was skipped
Averaging the channels cancelled a reversed one, so a real hit was reported
as silence. The skip warning also compared against the decode margin, so a
peak just before the kept window produced no warning.
* feat(cli): snapshot --keep redoes one board still and leaves the others
* fix(skills): the board recipe waits for a released snapshot --keep
* fix(cli): snapshot --keep refuses two times with one name and replaces linked stills
Miguel Ángel