Problem: computed Object Matte segmentations lived only in the in-memory
cache, so every restart or memory eviction segmented the shot again (After
Effects 26.5 keeps Object Matte results in its disk cache).
Fix: the persistent disk cache gets a third entry kind, Matte (folder
`mattes`, LZ4-compressed FrameSeg bytes, same checksum, atomic writes, LRU
eviction under the one size limit, purged with the rest). Entries are keyed by
the segmentation's chain key and scale; chain keys already hash the Input Key
(the layer's source, timing, masks and effects above), the prompt, the mode and
the model, so any change simply misses. The effects crate gets a SegStore
trait behind its memory cache: Chain::ensure reads from it on a memory miss
and writes what it computes. The engine installs the session's DiskCache as
the store (held weakly, so switching the disk cache off drops it). Roto Brush
segmentations go through the same chain and are persisted too. Frozen mattes
already persist in the project (the Frozen parameter).
Tests: engine tests_roto::object_matte_mattes_persist_in_the_disk_cache (a
second session on the same folder and project gets all tracked frames back
without being able to read a single input frame, renders identically, and a
moved prompt finds nothing); render disk_cache
segmentations_round_trip_and_reopen (round trip, reopen, damaged file is a
miss, size limit evicts).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem: Object Matte computed its frames only on demand while rendering, so
playing a shot stalled on every untracked frame, there was no progress or way
to stop, and nothing kept the mattes once computed.
Fix: Object Matte runs through Roto Brush's job machinery (one job slot,
cancellable with roto.cancel, a worker thread on the desktop, a Web Worker in
the browser, never the UI thread). objectMatte.propagate computes the frames
around the reference frame (clipped to the layer) outward, filling the same
content-keyed segmentation cache the renderer reads; the engine's sync now
queues an instance for background tracking when its frames or prompt change.
objectMatte.freeze stores the final mattes in a new hidden Frozen parameter
(Roto Brush's FrozenCache format) under a key of the input, prompt, span,
model and matte settings; renders use them without segmenting while the key
matches; a frozen prompt is refused until objectMatte.unfreeze. Freeze is one
undo step and is dropped if the prompt changed meanwhile.
objectMatte.status reports progress, frozen state and computed frames. Effect
Controls shows Freeze / Unfreeze and Track / Stop with a status line (progress
banner). The frontend now polls Roto Brush / Object Matte jobs on their own:
before, a finished job (and so a Freeze result) was only picked up while a
Warp Stabilizer job was active. Job progress is set before the worker starts.
Tests: engine tests_roto::object_matte_background_job_cancel_and_freeze
(queued by the edit, returns at once with progress, cancel keeps a partial
result, completion computes all 25 frames, frozen renders recompute nothing,
unfreeze recomputes, undo restores); effects object_matte job_frames and
frozen-matte tests (stored pattern used, key mismatch and damaged data
ignored); ui-egui ui_object_matte Effect Controls Freeze end to end.
docs/agents.md: Object Matte section; docs/effects.md regenerated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem: Object Matte could only be applied through the objectMatte.apply
command; After Effects 26.2 picks the object directly in the viewer (click it,
or drag a box around it).
Fix: an Object Matte tool (with Roto Brush & Refine Edge in the Alt+W slot,
icon drawn in code). In the Composition viewer a click sets a point prompt and
a drag a box prompt (its centre the point) on the footage, solid or comp layer
under the pointer, converted to layer pixels through the layer's transform.
The layer's Object Matte (the selected one, else its last) is re-prompted;
Shift adds another. Everything goes through objectMatte.apply, which gained
`effect` to re-prompt an instance (point, box or no box, reference frame; the
reach kept unless given), so each prompt is one undo step; the effect is then
selected, so its point controls show in the viewer. Hover draws a crosshair
and the outline of the layer that would get the matte (a live matte preview
would need a segmentation per pointer move). Automation ids:
viewer.objectMatte.crosshair / .target / .box, tools.ObjectMatte.
Tests: ui-egui ui_object_matte (hover, click, drag box re-prompts the same
effect, Shift-click adds one, undo steps); engine
tests_roto::object_matte_reprompt_updates_the_instance. Checked with the
headless snapshot example (hover crosshair, box drag isolating a disk).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem: the hand-maintained effect count in app.capabilities still said 306 after Object Matte was registered.
Fix: bump it to 307.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem: After Effects 26.2's Object Matte (pick an object with a click,
optionally a box, get a matte that follows it) had no counterpart; only
stroke-driven Roto Brush existed.
Fix: an Object Matte effect (Effect > Matte, ec.matte.objectmatte) turns
the Object Point (and optional box) on a Reference Frame into Roto Brush
strokes and segments them with Roto Brush's engine, content-keyed cache
and the model chosen in Settings > Roto Brush (MobileSAM, classical graph
cut without it). Track Object carries the matte frame by frame up to 600
frames each way; otherwise every frame is segmented on its own. Feather,
Contrast, Shift Edge, Invert and a Matte view. The engine keeps its Input
Key current so upstream edits recompute it. objectMatte.apply creates one
from agents / the CLI. Unusable prompts pass the layer through.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
file.import stores a path as given, so footage imported by the CLI or an
agent as `clip.mp4` was saved relative, and opening the project looked it
up from the working directory. The CLI, started in the project's folder,
found it; the desktop app, started from the Finder, Explorer or a launcher
in another folder, showed and rendered neither its picture nor its sound,
while the Project panel still listed the saved duration.
On open, a relative footage path that doesn't exist from the working
directory but does next to the project file is resolved there (After
Effects also looks for footage relative to the project). Paths that exist
as they are, and absolute ones, are kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Preserve legacy selection repair, modifier routing, viewer lifecycle and workspace persistence through the shared docking integration. Qualify the complete workspace gate by refreshing the 18 stale Audio Spectrum hashes for the existing linear renderer, with an independent synthetic-tone regression.
With several layers selected, After Effects applies a property edit to the same property of
every selected layer; we changed only the layer whose row was edited (the Properties panel
showed only the first selected layer), so a stopwatch switched on one layer and later moves
of the others had nothing to keyframe (#355), and scaling or rotating a selection changed
one layer (#491).
- Engine: `prop.set`, `prop.toggleAnimation` and `prop.toggleKey` take `selected: true`:
the change also goes to the same property (by match path, `transform/scale`,
`effects/<effect>#n/<param>`) of every other selected, unlocked layer whose value has the
same kind, in one undo step (a scrub keeps folding into it through its `merge` key). With
`offset: true` (a scrub) each layer changes by the same amount from its own value; a typed
value sets them all. Position with Separate Dimensions is handled per layer.
- The Timeline, Properties and Effect Controls panels pass `selected` (and `offset` while a
value is dragged). The Properties panel heads "(N) Selected Objects" (zh-CN / zh-TW / ja /
uk rows).
- Viewer: a bounding-box handle drag scales every selected layer by the same ratio about its
own anchor point; the Rotation tool turns them all by the same angle (one undo step).
Tests: engine test (stopwatch, scrub offset, typed value, keyframe button, one undo step,
unselected layers untouched); kittests: the Properties panel with three layers selected
(header, stopwatch on all three undone in one step, a scrub moving all three by the same
amount) and a handle drag scaling two selected layers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Measured against After Effects 26.3 with the original
examples/ae-spatial-tangents.jsx (three paths, ten keys, even and uneven
timing): an auto-Bezier key's out tangent is (next - previous) / 6 and
the in tangent its mirror, whatever the key times; an end key uses its
one neighbour, (neighbour - key) / 6. EffectCraft gave end keys none and
scaled middle tangents by segment length, so paths with auto-Bezier keys
ran up to 28 px from AE's. All ten tangents now match; path samples are
within 0.05 px.
Tests that relied on end keys having no tangent now expect AE's: Spatial
Bezier keeps auto keys' handles (linear keys still get a third), the
motion-path handle is drawn at a sixth, and a straight loopOut path is
compared within 1e-9. docs/fidelity records the method and scores.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The UI behaves as in After Effects (the Import dialog's Sequence checkbox
starts off, dropped files are stills, dropped folders make sequences), but
file.import without sequence or drag keeps importing a numbered still's
run, so existing scripts and agents keep working.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
After Effects' `app.project.workingGamma` (2.2 or 2.4, ignored with a working space) sets
the curve an unmanaged project uses: at 2.2 the piecewise sRGB curve, at 2.4 a pure 2.4
power curve. EffectCraft always used sRGB, so linear EXRs (Interpret As Linear Light) from
a 2.4 project came out darker in the shadows (0.18 → 0.461 instead of 0.490).
- `ProjectSettings::working_gamma` (default 2.2, serde default for old projects).
- The colour pipe's unmanaged curve is the Working Gamma's (sRGB, or Rec. 709's pure 2.4
curve): Interpret As Linear Light encodes untagged footage with it and Blend Colors Using
1.0 Gamma linearizes with it; it is part of the pipe's cache key. The GPU path takes the
same conversions.
- `file.projectSettings {workingGamma: 2.2|2.4}` (other values are refused, as in AE),
`app.project.workingGamma` in scripts and a Working gamma choice in Project Settings.
Test: render test with linear 0.051 / 0.18 / 0.5 footage (0.290 / 0.490 / 0.749 at 2.4,
0.250 / 0.461 / 0.735 at 2.2, the values measured in AE), linear blending at 2.4, a working
space ignoring the setting, and the cache key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
With no GPU compositor the report only said "GPU: none", though the host had recorded the
reason (`Session::accel_note`: the compositor's setup error, a retired GPU preview, a
headless start) and the window's wgpu adapter was known. Users with a capable GPU (#596, an
RTX 4070) couldn't tell why it wasn't used.
The report now carries `gpuNote` (the reason, when there is no compositor) and
`windowGraphics` (the adapter, backend, device type and driver the window draws with, which
the desktop app records when it sets up the GPU), and the text shows
"GPU: none (CPU compositing): <reason>" and "Window graphics: …"; the issue line names the
reason too.
Test: the compatibility report test sets a reason and an adapter and checks the JSON and text.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Cause: SequenceOptions::from_params defaulted `sequence` to true for
file.import, so any numbered still (IMG_0001.jpg) brought its whole numbered
run from its folder, and several picked or dropped stills with a common
prefix became one sequence. Dropped files never asked, and the Import
dialog's "<format> Sequence" checkbox started ticked.
Fix, matching After Effects: file.import takes picked and dropped files as
stills unless `sequence: true` (the Import dialog's Sequence checkbox, which
now starts unticked; Force Alphabetical Order implies it), while the numbered
stills inside a picked or dropped folder still import as sequences
(SequenceOptions::folders), and `sequence: false` (Alt-drag of a folder)
imports a folder's files one by one. Replace Footage, reload, proxies and
render outputs keep detecting their sequence (SequenceOptions::default and
from_params are unchanged), and mediaBrowser.import keeps bringing a numbered
still's run unless `sequence: false`. docs/footage.md and docs/agents.md
describe the new defaults.
Tests: sequence import_takes_stills_unless_asked_for_a_sequence and engine
numbered_stills_import_as_stills_by_default (one dropped still and two picked
IMG_ stills stay stills, a dropped folder makes one sequence, the checkbox
brings the run); the existing sequence tests pass `sequence: true`; ui
import_dialog_offers_the_image_sequence_option (unticked by default: a still;
ticked: the run) and a_dropped_folder_imports_as_one_sequence_with_sequence_interpretation
(two dropped frames: two stills; the folder: one sequence).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
(cherry picked from commit aa3c31db9ab3a774ee2f8ba867e1bafc1fce7aba)
Cause: render::content_bounds (used by snapping's targets and source, Align,
Distribute, the viewer's bounding box and handles) took shape layers' bounds
from draw_bounds, the rasteriser's safety rectangle: 1 px of anti-aliasing
padding around fills and width x miter limit / 2 + 1 around strokes. Two
squares snapped edge to edge therefore stayed 2 px apart, or much more with
strokes.
Fix: shapes::content_bounds uses visual bounds, as After Effects' layer bounds
do: the path bounds plus half the stroke width (through the draw's group
scale), with no anti-aliasing or miter padding. Rendering keeps draw_bounds.
Tests: render shape_content_bounds_are_the_visual_edges (a 100 x 100
rectangle is exactly +-50, +-55 with a 10 px stroke); engine
shape_layers_align_by_their_visual_edges (widths 100 / 110, Align Left puts
both edges exactly at 0).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
(cherry picked from commit bfd0abd7437ad1d38652e3aef51f9ba5dea9100b)
Cause: puppet::solve binds each moving pin to its nearest mesh vertex and fixes
that vertex at position + (vertex - rest), an offset in rest space with no
rotation or scale. puppet.addPin set a new pin's position to the click, so in
a bent region the bound vertex was pulled to where it would be unrotated, and
the mesh jumped as soon as the pin was added.
Fix: addPin picks the position that keeps the bound vertex at its current
deformed place: the vertex's deformed position minus the rest offset. On an
unbent mesh that is the click itself (round-off is snapped back to it), so
existing pins behave as before.
Test: a_pin_added_to_a_bent_mesh_moves_nothing (two pins, one moved to bend
the layer, then pins added at three places on the bent part: at most a few
edge pixels change from solver round-off; before the fix about 250 did).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
(cherry picked from commit 16151203b304d565b425873fdd33ae082cae483b)
Cause: edit.copy handled text editing, selected keyframes, effects and shape
items, then fell back to copying the selected layers. Selecting a property
selects its keys, so an animated property copied, but an unanimated one (or
the Transform group, whose keys prop.select doesn't select) copied the whole
layer, and Paste added a duplicate layer instead of setting the property.
Fix: with properties (or the Transform group) selected, Edit > Copy copies, as
After Effects does, each animated property's keys (the selected ones, else all
of them, through keys.copy) and each unanimated property's value into a new
value clipboard keyed by match path. Edit > Paste sets those values on the
same properties of the selected layers (a key at the CTI where the target is
animated) and pastes the keys at the CTI, as one undo step (keys.paste now
honours `merge`). Keys alone still answer with their count, as before.
Test: copy_a_selected_property_pastes_onto_the_same_property (static
Position, then the Transform group with animated Opacity: values and keys
land on layer B, no layer is added, one undo step each).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
(cherry picked from commit 8267519ebe20e889196e538de23b85b04a1fa76a)
A running script reached `prefs.set` through `app.run` and could turn on
Settings > Scripting & Expressions > Allow Scripts to Write Files and Access
Network (or turn off Warn User When Executing Files), which the File, Socket
and system.callSystem gates then read: the permission the user withheld was
the script's to take.
Commands a script issues now go through `Session::execute_for_script`, which
marks the session as running for a script for the command and everything it
runs (engine.batch, ScriptUI callbacks, scheduled tasks). While it is,
`prefs.set` refuses any change that would leave the Application Scripting
settings different (compared on the resulting settings, so dotted keys,
whole-page objects and batches are all covered), and a script's
`prefs.reset` keeps them as the user set them. Command scripts read from a
file (File > Scripts > Run Script File... with .json/.jsonl) count as
scripts too. The UI, CLI, control channel and MCP still change them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
parse_timecode reads bare digits as timecode pairs, so 00100 went to
frame 30 (1 s) instead of frame 100.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Progress docs per craftrules standards/progress-docs.md: target-app parity, gaps, roadmap with alpha gate, readiness by audience (full / mainstream / essentials) with Opus 5.5 hours, and feature, format, hardware, UI and localization parity docs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Timecode and Frames entries were parsed as absolute counts without the composition Start Timecode or Start Numbering Frames At offset, so committing the displayed text jumped to another frame.
A project saved with an inverted work area, or a comp shorter than a
frame, made Tick::clamp panic on the next start or end edit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
- file.relinkFootage (#461): close the params doc so checked execution
(MCP, control channel) rejects unknown keys; translate its menu entry.
- general.language lists Português (Brasil) (#217).
- docs/agents.md: set_property / add_keyframe take timeBase (#522).
- #299's MCP test no longer sends get_property a value.
- The Automatically Reload Footage sequence test uses PNG frames.
- h264_frames_at_reduced_sizes (#507): the bar depends on the fixture's
encoder; 50 dB with ffmpeg 6.1's x264.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
On an ASUS G513QY (Radeon RX 6800M, Adrenalin 26.9.2, Windows 11) every
launch died with STATUS_ACCESS_VIOLATION inside amdvlk64.dll, on a
driver worker thread, a moment after the first frame: eframe shows the
window it created hidden, the resulting resize re-creates the swapchain,
and the driver crashes there (same offset on two driver versions, with
the OBS/Steam/Epic Vulkan layers disabled, and with FIFO, immediate and
latency-1 presentation). DirectX 12 and OpenGL open the window fine.
A crash in a driver can't be caught, and the launch guard (#243) removed
its marker at the first frame, so this looked like a successful launch
every time and Window Graphics never switched. The marker now stays
until the window has kept drawing for SETTLE (3 s) or the app quits
normally before that; frame_drawn returns the time left and the app
asks for a repaint then, so an idle window settles too. The notice,
the Settings note and docs/preferences.md describe the new rule.
Verified on that machine with a scratch settings folder: launch 1
(Vulkan) crashed and left the marker, launch 2 switched to OpenGL with
the notice and closed with exit code 0, launch 3 stayed on OpenGL
without the notice.
Tests: effectcraft launch_guard
a_launch_that_stops_after_its_first_frame_switches_the_next_to_opengl
(a first frame then drop: the next launch gets GL and the notice; a
normal quit in that time leaves no marker), and the existing test now
asserts that the first frames don't settle a launch (marker present at
SETTLE/2, gone at SETTLE).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The params parser reads ':' as the start of a value, so 'OpenType?:
discretionaryLigatures?' declared a bogus OpenType key and hid
discretionaryLigatures, and 'ss01?…ss20?' declared only ss01.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Ported onto the current language plumbing next to Spanish (catalog codes,
Match System for pt-* locales, native menu Services label, docs, tests);
the language list and settings validation stay the shared LANGUAGES table.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Ported onto the current language plumbing next to Ukrainian (catalog
codes, Match System, native menu Services label, docs, tests), and the
stale Ease Presets row (a ScriptUI panel since #404) dropped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
Image sequences are watched only with All Footage (the default,
Non-Sequence Footage, leaves them alone). Purge also forgets the viewer's
per-revision disk keys, which hash footage files' size and time, so a
file reloaded without a new revision is keyed again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QiAJeFttZp83Fh7YuzAR5f
vangel76
skylar taylor-barrick
Claude
Katt
Brandon Thomas
MarekSpodBiedry
Nigel Ihza Firdausy
Vikram Bhaskaran
sanyamsingh001
Jeremy Yao