mirror of
https://github.com/storytold/effectcraft.git
synced 2026-11-06 18:41:26 +00:00
vangel76andClaude Opus 5.5
eedd64906c
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>
11 KiB
11 KiB
Effects parity
Last reviewed: 2026-10-12 · Last updated: 2026-10-12 · Change: minor (effect colour labels; earlier change: parameter count and closed issue references re-audited against
upstream/main470ea7a3and the generated effects.md) · Target: Adobe After Effects 2026
Effects are the deepest single area of After Effects. This page separates the four questions that a single "306 of 306" hides: does each effect exist, does it run fast, does it look like After Effects', and does the preset library exist. The per-effect catalogue is the generated effects.md; the overall numbers are in target-app-parity.md.
Summary
| Question | Value | Kind |
|---|---|---|
| Effects present | 306 registered, 3,017 parameters; all of After Effects' effects by name (298 at the M9.11 count) plus Obsolete and our own | Measured: registry, effects.md. After Effects 26.5's Plug-ins/Effects holds 306 .plugin bundles in 231 entries (one bundle can hold several effects, so this is a cross-check, not a match) |
| Parameter names, order, popups, units, defaults | aligned in M9.12 | Agent-checked against After Effects' documentation, not against a live dump |
| On the GPU | 280 (92%) | Measured: registry flags; GPU vs CPU agreement tested per effect |
| 32-bit float, no clamping | 306 | Measured: registry flags |
| Rendered output matches After Effects | partly measured | Some colour effects are checked against After Effects 2026 renders of our own test projects: Tint, Tritone, Black & White, Exposure, Levels, Curves, Luma Key, Extract, 4-Color Gradient and Fill within 1–3/255, Brightness & Contrast up to 6.2/255 (Color Correction, Generate and Keying rows below); Radial Blur, Glow, Venetian Blinds, Radial Wipe, Fractal / Turbulent Noise and Transform are checked the same way (rows below); every other effect is unmeasured. effectcraft_effects::fidelity::compare_frames exists for the corpus (fidelity/README.md); simulation effects are pinned by golden hashes of our own output (M13.27) |
| Animation presets | 76 original presets (8 text-animator, 23 Transitions, 23 Backgrounds, 22 Behaviors) against After Effects' 621 .ffx presets |
Measured: crates/engine/presets/text_animators.json, crates/engine/presets/transitions.json, crates/engine/presets/backgrounds.json, crates/engine/presets/behaviors.json; the After Effects Presets folder |
| 26.x additions | Curl Noise (26.3) present (#463 open); effect colour labels (26.5) present in Effect Controls only, not in the Timeline; the rebuilt Effect Controls (26.5) missing | Release notes |
Breadth ≈ 85% (presets weigh in; the effects themselves are ≈ 100%). Ready ≈ 40%: until a corpus compares renders, each effect is assumed to differ somewhere, and users keep confirming it.
By category (from effects.md)
| Category | Effects | On the GPU | Known differences or bugs |
|---|---|---|---|
| 3D Channel | 8 | 8 | Cryptomatte selection can't be set in the UI and the GPU path is slow (#473; quoted names with commas fixed); multi-layer EXR: one decode per frame for the picture and the channels, cache softly bounded at 1 GB (#474; holding channels as f16 is still to do) |
| Audio | 13 | 0 | CPU by design (applied in the mixdown) |
| Blur & Sharpen | 16 | 15 | Radial Blur keeps the layer edges as After Effects 2026 (713 now pixel-identical); Gaussian Blur interaction and output to be measured first (fidelity/ae-2026-acceptance.md) |
| Channel | 13 | 13 | — |
| Color Correction | 40 | 40 | Tint, Tritone and Black & White (sliders and Tint) match After Effects 2026 renders (Rec. 601 luminance); Levels (Individual Controls) matches After Effects 2026 in 32 bpc (channel controls before RGB, "Off for 32 bpc Color" unclipped; plain Levels' channel order not measured); Brightness & Contrast (non-legacy) close to After Effects 2026 ramps, not identical (up to 6.2/255 off; the test allows 7/255) (negative settings invert the positive curves); Curves interpolates a natural cubic spline through its points, as After Effects 2026 (exact on its grey steps); Tint has Swap Colors (#963); ACES 2.0 / OCIO 2.5 (26.5); Apply Color LUT (#514) and 32-bpc blending (#494) are closed |
| Distort | 39 | 35 | Transform matches After Effects 2026 renders (skew after rotation, drawn outside the layer); Liquify has no viewer brush; #303 (Mesh Warp drag) and #534 (CC Power Pin Unstretch) are closed |
| Expression Controls | 9 | 9 | — |
| Generate | 26 | 25 | 4-Color Gradient's Blend matches After Effects 2026 (weights 1 / (d² + Blend²)); Fill's Opacity is the filled layer's opacity, as After Effects 2026 renders it (with a Fill Mask it still mixes with the original, not measured); Fractal on the CPU (needs f64); #526 (Reset) and #536 (Write-on marks) are closed |
| Immersive Video | 12 | 12 | — |
| Keying | 12 | 12 | Luma Key and Extract key on Rec. 601 luminance, as After Effects 2026; Key Cleaner approximated; #533 (wire-removal Frame Offset) is closed |
| Matte | 6 | 4 | Mocha shape reads our documented JSON, not Mocha's format |
| Noise & Grain | 13 | 12 | Fractal / Turbulent Noise lattice matches After Effects 2026 (Scale 100 = 66.7 px cells); the noise values themselves differ (seeds) and Soft Linear contrast is about 20 % lower than AE's; Curl Noise (#463); Match Grain on the CPU |
| Obsolete | 9 | 9 | Luma Key matches After Effects 2026 on a grey ramp (hard cut, Tolerance only for Similar / Dissimilar) |
| Paint | 1 | 0 | Tablet pressure not read (hardware-parity.md) |
| Perspective | 10 | 9 | Chromatic Aberrations pan/tilt crash (#580) |
| Simulation | 18 | 18 | Particle Playground lacks Particle Exploder, text particles, ephemeral mapper; CC Particle World camera fixed (#397) |
| Stylize | 25 | 25 | Glow's halo (coverage of a dim layer's edge) and blur width match After Effects 2026 renders; glows above the threshold match too (each channel 8 × (value − threshold + 0.061), measured at three thresholds and two radii); Composite Original's labels are swapped against After Effects (EffectCraft's On Top adds the glow over the original, which After Effects calls Behind); Motion Tile output size (#369) |
| Text | 2 | 2 | — |
| Time | 8 | 8 | #535 (Pixel Motion Blur, Vector Detail) is closed |
| Transition | 17 | 17 | Venetian Blinds matches After Effects 2026 renders (band phase from the layer centre, Gaussian feather on both edges) except within ~6 px of the layer edge (up to 48/255); Radial Wipe's Feather matches After Effects 2026 (a Gaussian of σ = 0.4 × Feather pixels on both edges of the wedge) |
| Utility | 9 | 7 | — |
Third-party bundles in After Effects (Cineware, Mocha AE) are out of scope; Keylight is implemented under a generic name ("Key Light").
Estimate
≈ 70–130 h: a fidelity corpus case per effect at default and one non-default setting, compared with After Effects renders, and the fixes they turn up (40–80 h, most of it parallel, but every reference render needs After Effects on the owner's Mac); an original preset library covering After Effects' preset categories (Shapes, Synthetics, Text, Image effects remain; Backgrounds, Behaviors and Transitions exist) 15–35 h; the open effect bugs 10–20 h. Moving Fractal and the analysis effects to the GPU is performance, not parity.
Revision history
| Date | Change | Summary |
|---|---|---|
| 2026-10-19 | minor | Glow (original colours): bright pass per channel, 8 × (value − threshold + 0.061), measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | Glow (colour channels): the glow covers the layer's blurred alpha, blur σ = 0.384 × radius, measured against After Effects 2026 renders (CPU and GPU) |
| 2026-10-19 | minor | Radial Blur repeats the edge instead of sampling transparency outside the layer, measured against After Effects 2026 renders (CPU; Radial Blur has no GPU path) |
| 2026-10-19 | minor | Venetian Blinds: stripes measured from the layer centre, both band edges Gaussian (σ = Feather / 2), measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | Radial Wipe: Feather is in pixels on both edges of the wiped wedge (was degrees on the moving edge only), measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | Fractal / Turbulent Noise: lattice cell = Scale × 2/3 px (Block noise every 66.7 px at Scale 100, as After Effects 2026), CPU and GPU |
| 2026-10-19 | minor | Transform: After Effects' skew order and output beyond the layer bounds, measured against After Effects 2026 renders of three cases (CPU and GPU) |
| 2026-10-19 | minor | Tritone uses Rec. 601 luminance, measured against After Effects 2026 renders of a red solid and a grey ramp (CPU and GPU) |
| 2026-10-19 | minor | Black & White's Tint keeps the gray's luminosity (Color blend), measured against After Effects 2026 renders (CPU and GPU) |
| 2026-10-19 | minor | Levels: "Off for 32 bpc Color" no longer clips in 32 bpc projects, unclipped values extrapolate below Output Black; Levels (Individual Controls) applies the channel controls before the RGB ones, measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | Brightness & Contrast: negative Brightness / Contrast are the inverse curves of the positive ones (Brightness −40: 20/255 → 6/255 off After Effects 2026; still up to 6.2/255 off, not the same pixels), CPU and GPU |
| 2026-10-19 | minor | Luma Key: Key Out Brighter / Darker cut hard at Threshold, measured against After Effects 2026 renders of a ramp (CPU and GPU) |
| 2026-10-19 | minor | Luma Key and Extract: Rec. 601 luminance (was Rec. 709), measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | 4-Color Gradient: each point weighs 1 / (d² + Blend²), measured against After Effects 2026 at Blend 10 / 50 / 100 / 300 (CPU and GPU) |
| 2026-10-19 | minor | Fill without a mask: the layer becomes the fill colour at its alpha × Opacity (was a mix with the original), measured against After Effects 2026 (CPU and GPU) |
| 2026-10-19 | minor | Curves: natural cubic spline through the points, clamped to 0..1 (was a monotone spline), measured against After Effects 2026 (CPU and GPU); Lumetri's curves and Glow's Arbitrary Map keep the monotone spline |
| 2026-10-18 | minor | Tint uses Rec. 601 luminance weights, measured against After Effects 2026 renders of red, green, blue and grey solids (CPU and GPU) |
| 2026-10-18 | minor | Exposure converts to linear light with a pure 2.2 power curve, as measured in After Effects (Offset 0.0005 / 0.002 / 0.02 / 0.1 on black: 8 / 15 / 43 / 90), not the piecewise sRGB curve; its sliders take After Effects' ranges (Exposure −100..100, slider −4..4; Offset −2..2, slider −0.5..0.5; Gamma Correction 0.1..10) |
| 2026-10-12 | minor | Parameters 3,014 to 3,017 (effects.md); #303, #514, #526, #533–#536 marked closed instead of pending |
| 2026-10-12 | minor | Effect colour labels (26.5) added in Effect Controls |
| 2026-10-12 | minor | Tint's Swap Colors button and agent-drivable command (#963); undo/redo, instance targeting, animated layer-time edits and invalid-input tests |
| 2026-10-10 | major | First effects parity checklist |