The pool learned that an effect reads a footage item's channels only at the
first aux() call, so the first frame was decoded twice (picture, then in
full), and the hint was never cleared after the effect was removed.
The session now gives the footage source the whole set of files read by
enabled 3D Channel effects (FootageSource::set_aux_footage) with every
project change (Session::bump: edits, undo, redo, history jumps, rollbacks,
project open), sending it only when it changed. The pool keeps that set and,
as a safety net, what aux() learns; a new set drops what was learned.
Refs #474
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Review follow-ups:
- Files in use are never dropped from the channel cache: planes an effect
holds, the files of the last 4 pictures made of their planes until their
effects ask for the channels, and decodes in progress, which now count
the planes they will allocate (from the header) against the budget. With
4 frames at once over a budget of one file, each file decodes once (test).
- A decode that panics removes its entry instead of leaving it decoding.
- One decode allocates at most 8 GiB of planes (512 MiB on 32-bit and
wasm); a header asking for more is refused with the reason.
- A shared picture whose file has no cached channels (a failed read, a
broken part it doesn't use) is decoded from its own channels.
- Comment why waiting on a decode may block a rayon worker; drop the
subsampling check and test (the exr crate refuses subsampled files).
- Docs: the cache is a soft bound; a single-part file still decompresses
every channel; the first frame and the image crate's route decode twice.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds tiled (ZIP16, RLE), tiled mip-mapped (ZIP16, PIZ), decreasing line
order, uncompressed, a uint R/G/B picture and a multi-part file whose first
R/G/B part is depth-named (the planes can't stand in for the image crate's
picture there). The new hashes were taken with main's crates/media/src at
a646bdcc (extracted with git archive) and the same test, before the decode
rework.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The channel cache decodes a file once however many callers ask for it at
the same time (the others wait for that decode), and is bounded by the
bytes of its planes (1 GiB, least recently used out first, a file counting
as used when its decode finishes, the newest file always kept) instead of
four files of any size.
- Once an effect has read a footage item's channels, its pictures are made
of the cached planes, also for files with an unnamed RGB layer (bit for
bit what the image crate shows, which still decodes the files the planes
can't stand in for); footage alone decodes only its picture's channels
(a multi-layer file's were all decoded and cached for every frame).
- Channels are decoded straight into their display-window planes, and the
picture is converted from the planes without two interleaved copies.
Measured on a 30-frame 2560x1440 sequence of 32 channels (DWAA), one frame
at a time: footage alone 0.185 -> 0.073 s per frame and 4.4 -> 1.8 GB peak;
with an unnamed RGB layer plus Cryptomatte 0.244 -> 0.159 s; multi-part
plus EXtractoR 0.194 -> 0.127 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Synthetic single-layer, multi-layer, multi-part and windowed files in
half, float and uint, with RLE, ZIP, PIZ and DWAA, decoded through the
media pool picture first and channels first. The hashes are taken on
main before the decode is reworked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Follow-up to #2823: box each of several objects around its whole matte on correction frames
With several objects, a correction frame's box came from the model's cut
of that object's strokes on the frame alone. A stroke on part of the
object boxed only that part, and SAM 2 followed only that part from
there on. Each object's matte is now carried to the frame with its own
strokes, from its first prompted frame or its nearest tracked frame, by
the walk Chain::ensure uses (now Chain::carry). These mattes are not
cached, since object keys hold tracked results.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Follow-up to #2823: carry corrections before the base from the base outwards
Prompt frames were visited in ascending order, but before the base the
carry walks towards it, so each correction there re-walked from the
root. Visit each side from the base outwards and sort the prompts back
into frame order for the tracker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Follow-up to #2823: find an object's carry root in one pass over its strokes
The root search scanned every frame's strokes from the prompt frame to
the base, for every prompt. Take the object's foreground stroke on the
way that is nearest the base in one pass instead; the root is the same.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M13.35: reproduce an earlier object correction losing tracked matte
* M13.35: carry earlier object corrections from valid tracked mattes
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: FredApps <FredApps@users.noreply.github.com>
* Fix#2823: Propagate boxes SAM 2 around the matte shown on the stroked frame
The tracker's box prompt came from the classic graph cut of the strokes,
which can leak into the background; SAM 2 then filled the whole box.
Build the box from the model's own matte (segment_with), the one shown
before Propagate. segment_with falls back to the classic cut when the
model can't segment, so non-model paths are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Fix#2823: box a lone object around the matte shown on correction frames too
A correction frame shows a matte propagated from the frame before, not a
fresh model cut, so the tracker's box came from a different matte than
the one on screen. With one object the box now comes from Chain::ensure,
the matte shown on that frame. Several objects share one shown matte, so
each keeps the model's cut of its own strokes. Cancelling while the shown
matte is computed stops instead of falling back, and the prompt frame is
read only when needed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Under load its thread pulled samples before the test read the clock, so
'from the first frame' saw frame 1 and a_started_preview_follows_its_sound flaked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The PR pinned the head of a fork branch, not in craft-libs history; the crate
source is byte-identical at both revs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ProRes renders had no colr atom, so the stream reported no primaries,
transfer or matrix. Write an nclc atom, as FilmCraft does for MOV, with
the codes of what the render path makes: the project's output space when
it has a working space, BT.709 otherwise.
sRGB output stays at transfer 1, which is still open on #340 and #464.
The frames don't change, so the pinned ProRes bytes move by exactly the
atom's 18 bytes.
The craft-ecproj-format pin moves to storytold/craft-libs#409's new
head, where Head::read is the one reader and saved_by() is removed.
effectcraft-project's saved_by reads savedBy with Head::read, and from
ReadError::TooNew for a project with a newer schema.
saved_by now returns None for text that is not a JSON object with a u32
schema. EffectCraft calls it only on files Project::from_json opened,
so opening and importing projects are unchanged.
effectcraft-project writes and checks the envelope of .ecproj files with
craft-ecproj-format from craft-libs: File (savedBy, then the saved view
and the project) with to_json and to_json_compact for Save Compact
Project Files, check_schema and its message, SCHEMA_VERSION and
saved_by, which stay re-exported under their names. Project, SavedView
and the schema 1 conversion stay here.
Saved files are byte-identical. saved_by now reads only a JSON object,
so a project written as a JSON array no longer reports a savedBy.
A `trailer` marker inside an unreferenced stream followed by `(` and more
than 4 KiB of text made the lexer stop at the window's end, so a later
`/Encrypt` name anywhere in the file refused a plain file. The unbounded
scan read a string there and opened it. The past-the-window fallback now
applies only when no object starts in the window or the first one opens
a dictionary (`<<`, including one split by the window's end).
Regression test with `trailer(` and `trailer[` in a stream and a later
Encrypt name.
The pre-scan window could end inside a trailer key, so a decoy like
`/Encrypted` cut right after `/Encrypt` was taken for the Encrypt key
and an unencrypted file was refused. A window that ends mid-token now
ends before that token instead, and the past-the-window fallback reads
the name whole (a real `/Encrypt` cut as `/Encryp|t` is still refused).
Tests for that boundary, for a trailer with a long /ID, and for `#`
without two hex digits staying as written in a name.
vangel76andClaude Opus 5.5
Brandon ThomasandClaude Opus 5.5
FredApps
Brandon Thomas
dotCooCoo
JonRC
Jonas Neubauer
jayzed80
Dylan M. Taylor
tylerjenningswandChewybone-sudo
Patrik LermonandClaude Opus 5.5
becauseno7