Merge upstream/main into aep-import-quickwins

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/aep-format.md

@@ -328,7 +328,7 @@ Property record (`LIST tdbs`: `tdsb`, `tdsn`, `tdb4`, then `cdat` or a `LIST lis
 | keyed spatial-template flag | `tdb4` | 79 | u8 | 1 on a static colour or point effect value, 0 once it is keyed (359 / 364, 369 / 382); 0 on keyed and static sliders (362 / 363, 391 / 397); meaning **?** | 359, 362-364, 369, 382, 391, 397 | V (values) |
 | expression flags | `tdb4` | 76, 77, 120 | u8 | all three are 1 when an expression exists. 76 and 77 are also 1 on a camera's or light's computed (auto) Position, which has no expression | 060, 213, 251 | V |
 | **expression disabled** | `tdb4` | 119 | u8 | 1 = the expression is disabled; the text and bytes 76, 77, 120 stay. On the built-in view layers the byte flips by itself after the first save, so it counts only together with an expression | 060 (0), 213 (1) | V |
-| state | `tdsb` | 0 | u32 | 1 = written (non-default value, keys or expression), 3 = default state | 040, 071, 073 (1); 3 on the unwritten properties of 000 | V (1), E (3) |
+| state | `tdsb` | 0 | u32 | 1 = written (non-default value, keys or expression), 3 = default state. **A layer, camera, light, shape or text property in the default state holds After Effects' default** or a value with no effect on that layer (survey of 839 corpus and sample files: Material Options, cone 90 / 50 and falloff 500 / 500 of point lights, Position `[0, 0, -1333.33]` of ambient and environment lights, Opacity 1 and Scale `[1, 1, 1]` of cameras and lights, a polygon's Inner Radius 50, Miter Limit 4, the inert Time Remap 0), so the importer skips it. **An effect parameter's record in the default state holds the effective value** (3.7, stored value) | 040, 071, 073 (1); 3 on the unwritten properties of 000; 1200, 1400 (effects) | V (1), E (3), V (effects) |
 | static value | `cdat` | 0 | `components` x f64 | the first `components` doubles are the value | 040, 041, 043, 071 | V |
 | user name | `tdsn/Utf8` | | text | `-_0_/-` means "no custom name" | 091 (`Mask 1`), 238 (`Group 1`) | V |
 
@@ -437,9 +437,9 @@ The expression engine is project-wide (`LIST ExEn/Utf8`, `javascript-1.0`). API:
 | **float slider valid minimum, slider range** | `pard` | 104; 112 / 116 | f32; f32 x 2 | the smallest valid value (Exposure -100, Offset -2, Gamma Correction 0.1 as f32) and the range the slider covers (Exposure -4..4, Offset -0.5..0.5, Gamma 0.1..10); the slider range is repeated as `tdum` / `tduM` (f64) in each value record (391-394). @120 holds the default again as f32 (Gamma 1.0, Blurriness 25), u16 @124 the display decimals (**H**: Exposure 2, Offset 4) | 389, 391-394 | V |
 | **definitions of a second instance** | top-level `LIST EfdG` | | | one `EfDf` per effect TYPE (`tdmn` = match name, then a copy of ONE instance's `sspc`, including that instance's value records); the `EfDC` count before them is a little-endian u32. When two layers carry the same effect type, only one instance keeps a full `parT`; **the other has an empty `LIST parT` (4 bytes)** but its own `tdgp` (input layer `-0000` with its own `tdpi`, the button record, and its value records). Its definitions come from the `EfdG` copy by match name, its values from its own records, never from the copy (383: the copy holds S100's Start [0.1, 0.2]; L's Ramp reads the 368 defaults). Which instance keeps the full `parT` is **?** (in 383 S100 is both first in the file and newest). The `EfDf` order follows the stack order (365 / 366); `Pefl/pjef` does not | 365, 366, 383 | V |
 | popup items | `parT/pdnm/Utf8` | | text | items separated by `\|` | 077 (`Item 1\|Item 2\|Item 3`), 214 | V |
-| stored value | `sspc/tdgp/<match>-NNNN` | | `tdbs` | present only for a parameter that differs from its default; `tdsb` 1, `cdat` as in 3.4. **Absent record = the `pard` default** | see below | V |
+| stored value | `sspc/tdgp/<match>-NNNN` | | `tdbs` | present only for a parameter that differs from its default; `tdsb` 1, `cdat` as in 3.4. **Absent record = the `pard` default.** Some parameters are recorded in the default state (`tdsb` 3) and the record still holds the value After Effects uses: every one of the 25 such records in the corpus and the samples differs from its `pard` default and reads back as recorded (Lumetri Color's hidden controls High Dynamic Range, HDR Specular, the second Look, Legacy Tone And Saturation, Face Detection, Auto Color, Auto Intensity and SemanticAutoTone Progress, 1200; Cryptomatte's Selection Mode on, 1400). No effect record in the default state holds a value that does not count. The reader takes the value in either state (`EffectParam::recorded_value`, `EffectParam::value`); `EffectParam::stored_value` keeps to written records, and a record in any other state (no sample has one) is not read: the importer reports it as damaged data. That other effects record values so is H (no sample shows one) | see below; 1200, 1400 | V (1200, 1400), H (other effects) |
 
-Stored values by type (all `tdsb` 1):
+Stored values by type (`tdsb` 1, or 3 as above):
 
 | Type | Stored | Evidence | Conf. |
 |---|---|---|---|
@@ -837,7 +837,7 @@ another value).
 **Masks, shapes**
 
 - Masks: locked, colour, feather falloff, rotobezier, keyed expansion, a path whose last vertex has only handles, the counter after a mask is deleted, a mask on a scaled layer (whether the box stays layer-relative), an animated path with three or more keys.
-- Shape layers: open shape paths (in shape groups), stroke dashes, taper and wave with values, merge paths, repeater, offset paths and the other path operators, item blend modes and composite order, shape direction, gradient colour stops other than the default (**?**: they appear only as an `ADBE Vector Grad Colors` entry), highlight length and angle other than 0, gradient scale and rotation; why a radial gradient writes its highlight records with `tdsb` 1 and a polygon its inner radius with `tdsb` 3 (**?**).
+- Shape layers: open shape paths (in shape groups), stroke dashes, taper and wave with values, merge paths, repeater, offset paths and the other path operators, item blend modes and composite order, shape direction, gradient colour stops other than the default (**?**: they appear only as an `ADBE Vector Grad Colors` entry), highlight length and angle other than 0, gradient scale and rotation; why a radial gradient writes its highlight records with `tdsb` 1 and a polygon its inner radius with `tdsb` 3 (**?**; the polygon's records hold the defaults 50 and 0, as every non-effect record in the default state does: 3.4, state).
 
 **Text**
 

docs/architecture.md

@@ -409,7 +409,12 @@ inputs in 32 bpc. With a working space (sRGB, Rec. 709, Rec. 2020, Display P3; m
 from the published primaries), footage is converted from its colour profile (stream metadata or
 Interpret Footage, else sRGB) and the top-level comp is converted to the sRGB display. Linearize
 Working Space runs everything in linear light; Blend Colors Using 1.0 Gamma linearises only for
-blending. Layer cache keys include these settings.
+blending. Layer cache keys include these settings. Project Settings ▸ HDR (compand / tone
+mapping) is the last top-level step, for standard-dynamic-range displays and outputs; OpenEXR
+renders (sequences and ProEXR) skip it (`RenderOpts::skip_hdr_step`) and decode the frame with the
+output space's own curve (`Pipe::output_linear`), so they hold linear light in the output space's
+primaries; unmanaged projects decode with the Working Gamma's curve (sRGB at 2.2, a pure 2.4 power
+curve at 2.4). Unmanaged projects and Rec. 2100 PQ / HLG outputs never get the HDR step.
 
 **Frame blending**: footage and precomp layers whose source time falls between source frames
 (rate conform, stretch, remap) blend the two neighbouring frames, by cross-fade (Frame Mix) or by

docs/effects-parity.md

@@ -26,7 +26,7 @@ corpus compares renders, each effect is assumed to differ somewhere, and users k
 
 | Category | Effects | On the GPU | Known differences or bugs |
 |---|---:|---:|---|
-| 3D Channel | 8 | 8 | Cryptomatte's GPU path is slow (#473; the Selection is picked in the viewer with an eyedropper, quoted names with commas fixed); multi-layer EXR decode shared (#474 open: f16 storage and cache sizing remain) |
+| 3D Channel | 8 | 8 | Cryptomatte's GPU path is slow (#473; the Selection is picked in the viewer with an eyedropper, 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](fidelity/ae-2026-acceptance.md)) |
 | Channel | 13 | 13 | — |

docs/file-format-parity.md

@@ -1,6 +1,6 @@
 # File format parity
 
-> **Last reviewed:** 2026-10-19 · **Last updated:** 2026-10-19 · **Change:** `.aep` / `.aepx` import (`crates/aep`: `container`, `import`) · **Target:** Adobe After Effects 2026
+> **Last reviewed:** 2026-10-19 · **Last updated:** 2026-11-05 · **Change:** ProRes alpha final-run conformance; previously `.aep` / `.aepx` import (`crates/aep`: `container`, `import`) · **Target:** Adobe After Effects 2026
 
 Every format After Effects reads or writes, with EffectCraft's support. After Effects' list comes
 from its installed bundle (26.5.0.89: `Info.plist` `CFBundleDocumentTypes`, `Plug-ins/Format`,
@@ -41,7 +41,7 @@ trips and renders, checked by tests, not against After Effects), **partial** (na
 | PNG | R/W | yes | yes (sequence) | good |
 | JPEG | R/W | yes | yes (sequence) | good |
 | TIFF | R/W | yes | yes (sequence) | good |
-| OpenEXR (multi-layer, Cryptomatte) | R/W | yes | yes (32-bit sequence) | good; HTJ2K compression fails (#482); EXRs without an unnamed RGB layer decode twice (#474) |
+| OpenEXR (multi-layer, Cryptomatte) | R/W | yes | yes (32-bit sequence) | good; HTJ2K compression fails (#482); a multi-layer frame decodes once for its picture and EXtractoR / Cryptomatte from the first frame on, except Media Replacement items, which are learned at their first frame, and files whose picture the `image` crate still decodes (#474); footage alone keeps only its picture's channels, though a single-part file still decompresses every channel's pixels; decoded channels are held as f32 |
 | Photoshop PSD / PSB | R (footage, composition) / W (layers, sequence) | yes (layers, groups, masks, styles, smart objects) | Save Frame As ▸ Photoshop Layers; no Photoshop sequence output | good, from Adobe's published spec |
 | Illustrator `.ai`, PDF, EPS | R | yes (vector, Continuously Rasterize, Create Shapes; Composition and Composition - Retain Layer Sizes, #1536) | — | partial: no JPX / JBIG2 images, no Illustrator procset EPS |
 | SVG | R | yes | — | good; SVG markup on the clipboard pastes as a shape layer (26.3, #738); Illustrator's own clipboard data (PDF/AI) doesn't |
@@ -62,7 +62,7 @@ trips and renders, checked by tests, not against After Effects), **partial** (na
 | MP4 / MOV with H.264 | R/W | yes | yes (own encoder, + AAC) | decoder bit-exact on conformance streams (FilmCraft) |
 | HEVC (H.265) | R/W | yes | yes (MP4 `hvc1`, Main / Main 10) | encoder bit-exact with ffmpeg's decode; no B-frames / SAO, larger files |
 | AV1 | R | yes | yes (MP4 `av01`, WebM) | beyond After Effects' export; no B-frames / CDEF |
-| Apple ProRes (422 family, 4444) | R/W | yes | yes | good; 4444 colour tag wrong (#340); no ProRes RAW |
+| Apple ProRes (422 family, 4444) | R/W | yes | yes | alpha final-run syntax checked in exported MOVs (#293); 4444 colour tag wrong (#340); no ProRes RAW |
 | WebM / Matroska (VP9, AV1, Opus) | R (limited) | yes, incl. VP9 alpha | yes | beyond After Effects |
 | MXF (OP1a / OP-Atom) | R/W | yes (FilmCraft's demuxer: AVC, MPEG-2 incl. XDCAM / D-10, VC-3 / DNxHD, ProRes; PCM audio) | none | partial: import (#767), checked against ffmpeg; no MXF output |
 | AVI | R | yes (FilmCraft's demuxer: AVI 1.0 / OpenDML; Motion JPEG, H.264, HEVC, uncompressed; PCM, MP3, MP2, AC-3 audio) | none | partial: import (#1701), checked against ffmpeg; H.264 needs its parameter sets in the chunks (global headers and copies with an `avcC` record don't decode until storytold/filmcraft#2303), and a damaged video index loses the video (storytold/filmcraft#2302); DV, MPEG-4 Part 2 (DivX / Xvid) and other legacy codecs import but show why they have no picture; no AVI output |
@@ -146,6 +146,7 @@ relinking 6–12 h, `.mogrt`-compatible templates 15–30 h (partly parallel wit
 
 | Date | Change | Summary |
 |---|---|---|
+| 2026-11-05 | minor | ProRes encoder pin and export regression for final alpha runs; native macOS decode validated, Resolve still unverified (#293) |
 | 2026-10-30 | minor | Render-queue source-protection behavior and limits consolidated here from the standalone page |
 | 2026-10-12 | minor | #324 and #514 marked closed; open `.aep` PRs #251 and #860 noted |
 | 2026-10-10 | major | First file format parity document, from the After Effects 26.5 bundle and this repository's import/export registries |

docs/footage.md

@@ -120,6 +120,15 @@ Nuke's `beauty`) show the beauty pass: the layer named Combined, Beauty, Image,
 Color, else the first colour layer that isn't a data pass. Cryptomatte, depth, mist, normals,
 vectors, positions and IDs are never the picture. EXtractoR reads every other layer and channel.
 
+Footage on its own keeps only the channels of its picture (a single-part file still
+decompresses all of its pixels to read them). Once an effect reads an item's channels
+(EXtractoR, Cryptomatte, the 3D Channel effects), each frame, the first included, is decoded
+once for the picture and the effects together, and kept for a while: about 1 GB of decoded
+channels (512 MB in the browser), a soft bound, as the newest file and files still in use
+stay whatever their size. Several frames asking for the same file at once wait for one
+decode. Once no enabled effect reads the item any more, its frames go back to the picture's
+channels alone.
+
 The Cryptomatte effect keys objects by their Cryptomatte ID passes, with the names from the
 file's manifest. Pick them with the eyedropper next to its Selection, then click in the
 Composition viewer (Shift adds, Alt or Ctrl removes).

docs/gaps.md

@@ -65,7 +65,7 @@ are grouped by area; the issue tracker has the rest.
 | 6 | **Linux and Windows interaction bugs** | Open: #488 rename arrows (hardening landed, reporter asked for details), #492 anchor drag, #268 file drops on Wayland (winit), #613 Direct3D 12 compositor. Now closed: #491, #493, #467, #320, #302, #596 | Many reports come from Linux; some Windows GPUs fail to start the compositor | 25–45 h | G2 · [ui-parity.md](ui-parity.md) |
 | 7 | **Animation presets: 76 against 621** | After Effects ships 621 `.ffx` presets (Behaviors, Backgrounds, Shapes, Synthetics, Text, Transitions…); we ship 76 original presets (8 text-animator; 23 Transitions: wipes, clock and fan reveals, blinds, dissolves, card flips, CC wipes, fades, #757; 23 Backgrounds: drifting gradients, fog, cloud and smoke, plasma and magma, cells, hexagons, scrolling grids and checks, pulse rings, bokeh, light sweep, #1096, driven by expressions so they keep moving, not seamless loops; 22 Behaviors: wiggle, shake, float, orbit, bounce, drift, pendulum, spin, pulse, heartbeat, blink, strobe, flicker on transform and opacity, #1510) and read only our `.ecpreset`. Missing: Shapes, Synthetics, Image and more Text | Presets are how many designers start; tutorials assume them | 15–35 h (original presets; `.ffx` reading is out of clean-room scope) | G8 · [effects-parity.md](effects-parity.md) |
 | 8 | **Mask and shape bugs** | Open: #1004 mask vertex multi-select, #999 auto-pan, #1042 / #1033 vertex nodes under the bounding box, #959 / #960 shape group toggles and operators. Now closed: #510, #513, #527, #509, #303, #553, #932 (Convert To Bezier Path on a shape layer's parametric path) | Masks are core compositing | 10–20 h | G2 · [target-app-parity.md](target-app-parity.md) |
-| 9 | **Effect behaviour bugs** | #580, #463 Curl Noise, #369 Motion Tile output size, #943 Turbulent Displace Evolution, #473 Cryptomatte selection (quoted names with commas fixed, 3b7f1ace; the viewer eyedropper added; GPU time open), #474 multi-layer EXR (decode shared, faffad3a; f16 storage and cache sizing open). Closed: #526, #514, #533–#536 | Each one breaks a look someone built | 10–20 h | G1/G2 · [effects-parity.md](effects-parity.md) |
+| 9 | **Effect behaviour bugs** | #580, #463 Curl Noise, #369 Motion Tile output size, #943 Turbulent Displace Evolution, #473 Cryptomatte selection (quoted names with commas fixed, 3b7f1ace; the viewer eyedropper added; GPU time open), #474 multi-layer EXR (one decode per frame for picture and channels, byte-bounded cache; f16 storage still to do). Closed: #526, #514, #533–#536 | Each one breaks a look someone built | 10–20 h | G1/G2 · [effects-parity.md](effects-parity.md) |
 | 10 | **Colour pipeline differences** | #340 ProRes 4444 tagged bt709 but sRGB-encoded (open); ACES 2.0 / OCIO 2.5 configs (26.5) missing; now closed: #494 32-bpc blend modes, #495 and #750 Working Gamma 2.4, #546 / #547 viewer colour | Wrong colour in delivery is a client-visible failure | 15–30 h | G1 · [file-format-parity.md](file-format-parity.md) |
 | 11 | **Text rendering and layout** | #453 CJK text in comps: the Han fallback now reaches the installed Chinese faces; open: #316 paragraph alignment inverted (maintainers' tests show Left/Right correct; asked for a project), #417 soft at Fit (mostly fixed by #444), #550 toolbar font options. Closed: #483, #583 | Text is in almost every motion graphic | 10–20 h | G2 · [ui-parity.md](ui-parity.md) |
 | 12 | **3D interaction and depth** | No 3D transform gizmos (#392, #347); camera aperture (#260); reflections (#261); extruded surface materials beyond Front / Bevel / Side / Back Color (fb5a5e9a; #315 closed); #307 reopened (the viewer's renderer button now shows and switches the renderer, #645); Z position (#489) closed | Advanced 3D is After Effects' growth area since 2023 | 20–35 h | G2 · [ui-parity.md](ui-parity.md) |