[vanceingalls]

What this adds

A new root attribute, data-motion-blur-phase, that sets when the shutter opens. It is in degrees of a frame, like the shutter angle, and clamps to -360 to 360.

The sampling window starts at t + phase / 360 / fps and lasts shutter / 360 / fps. So at 180°:

• unset, or -90: centred on the frame, as today
• 0: opens on the frame and runs half a frame ahead
• 360: opens one whole frame later

It works on both the capture and filter tiers, which now take the window from one function, shutterTimes in frames.ts. It is parsed in motionBlurSettings.ts, and the Studio panel's level control leaves it alone. Documented in docs/guides/motion-blur.mdx and the CLI docs.

Why

Page blur can only centre the shutter, and After Effects has a Shutter Phase setting. In a test corpus of 68 After Effects exports, 133 of the 134 blurred layers use a non-centred shutter (180/0 is the most common, then 299/360), and every one of those layers that moves in 2-D is off-centre. Without this, a smear lands in the wrong place, and the difference grows with speed. The After Effects exporter needs it to reproduce those shutters.

What it does not change

• Any existing page. With no attribute the window is exactly what it was: phase defaults to minus half the shutter, and the arithmetic comes to the same number. Renders below match the released 0.8.223 byte for byte.
• 3-D movers. Page blur still skips them.
• The engine's sub-frame blur, the shutter default, and the samples, tier and events attributes.
• The Studio controls. There is no UI for phase; an authored value survives a level change.

Evidence

A 2-D box moving 10 px/frame, shutter 180°, rendered through this branch. At t = 1 s the box's left edge is at x = 300. Smear columns give the first pixel with any ink and the first fully lit one:

phase	filter tier	capture tier	filter smear	capture smear
unset	df1341e3	b1e8ae7f	298 → 302	298 → 302
-90	df1341e3	b1e8ae7f	298 → 302	298 → 302
0	e471c1d8	88e006f4	300 → 305	301 → 305
360	2aaa32e7	85d1b52b	310 → 315	311 → 315

Unset and -90 are md5-identical to each other and to the same page on released 0.8.223.

Tests

New tests were seen failing first, and reverting the one line in frames.ts turns 4 of them red.

• Settings: reading, clamping at ±360, 0 kept as a real value, unreadable treated as unset, round trip.
• frames.test.ts: no phase reads exactly the old times; phase 0 at 180° starts at t; phase 360 starts a frame late; -90 at 180° equals unset; reads never go before 0.
• index.test.ts: a root attribute reaches the seek times on both tiers, and both stay centred without it.

Local: core 2,730 passed, Studio panels 95 passed; tsc, runtime typecheck, oxlint, oxfmt, comment citations and the comment ratchet all clean.

🤖 Generated with Claude Code

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
hyperframes	🟢 Ready	View Preview	Nov 2, 2026, 5:33 AM

💡 Tip: Enable Automations to automatically generate PRs for you.

[github-actions]

Edit accuracy: accurate 267 (base branch 267), smooth 94 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (8)

• seqplainkeys-none-center-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-center-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-center-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-pct-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-pct-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-px-r30-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run

Unstable (1)

• seqplainkeys-none-px-r0-root-z100: tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true

[jrusso1020]

Reviewed at 51da80f8. The code is good and I found nothing blocking. I'm posting COMMENT, not APPROVE, only because this head conflicts with main and CI never ran on it. I'll approve the rebased head once it's green.

Merge state. git merge-tree --write-tree HEAD origin/main shows two conflicts, both in tests:

• frames.test.ts is add/add. #7080 created the same file on main, with its own laidOut and rect helpers. Keep both describe blocks.
• index.test.ts gets your new describe at the end of the file, and main changed the same tail.

The source changes merge cleanly: frames.ts, filterTier.ts and index.ts all auto-merge, and #7080's readMoverFrames changes keep your start = t + req.phaseSeconds line intact. Every CI job at this head shows skipping, so there's no test evidence at this SHA yet.

What I checked

• shutterTimes is now the only place either tier computes the window (frames.ts; index.ts:277, filterTier.ts:189). I grepped shutterDeg, shutterSeconds and shutter / 2 on main, and no other site builds a window. The remaining shutterDeg uses are report fields (index.ts:563,703, filterTier.ts:231) and the Studio level presets.
• Unset is unchanged: -shutterSeconds / 2 is the old t - shutter / 2 term exactly, so no float drift. That fits your md5-identical renders.
• It matches the engine's sub-frame blur convention: windowStart = t + shutterPhase / 360 / fps, with default -90 at 180° (engine/src/services/motionBlur.ts:14,61,93). So phase means the same thing on both blur paths.
• Mover discovery (discoverMovers) scans the whole timeline, not the shutter window, so an off-centre window can't hide a mover from the scan.
• Studio round trip: settingsOf reads phaseDeg and changedEdits writes it back, and patchForLevel only touches enabled, events, shutter and samples. So an authored phase survives a level change, as the body says.

Nits (none block)

• Report has shutterDeg but no phase, so an off-centre render looks the same as a centred one in the [motion-blur] log. Adding phaseDeg beside shutterDeg in both tiers' report() would make it debuggable.
• At phase 360 with shutter 720 the window ends 3 frames past t, so the last frame seeks past the composition's duration. Seeks clamp to the end today, and a centred 720 already reads 1 frame past, so this is only more of an existing case. One test at t = duration would pin it.

Need: yes, the After Effects exporter needs non-centred shutters · Size: ok · Reuse: same phase convention as engine motionBlur.ts, one formula per package, sharing across the core/engine boundary isn't worth it · Simpler: already minimal · Structure: fine
Risk: low. Unset is byte-identical, and the new behavior only applies when the attribute is set.

Verdict: COMMENT
Reasoning: The change is correct and minimal. It can't be approved at this head because it conflicts with main and CI didn't run. I'll approve after a rebase and green CI.

— Rames

[vanceingalls]

Rebased onto main (the two test conflicts were only in frames.test.ts, kept both sides, and index.test.ts). Took the first nit: Report now carries phaseDeg. Not changed: phase 360 with shutter 720 reading past the end on the last frame. It follows the same convention as the engine's sub-frame blur and the old centred shutter also read past the end by half a shutter; say if you want it clamped. Core blur unit tests: 217 pass. @heygengenesis review please, and Rames, ready for your approve once CI is green.

[heygengenesis]

Verdict: REQUEST CHANGES · 1 blocker · 0 important · 0 nits · head 08896be

The code is correct, but the new Report.phaseDeg doc comment turns the required Comments check red. The fix is one line, inline.

Scope and CI

Additive to @jrusso1020's review at 51da80f8, whose checks I re-ran against the rebased head. shutterTimes is still the only place either tier builds the window, and the source merged cleanly. Both conflicted test files keep #7080's blocks next to the new shutter phase describes. phaseDeg is now on every Report built by both tiers and offReport, which addresses the nit.

CI at this head: CI, Docs, CodeQL, Windows, Player and the captures all pass. Two checks are red. Comments is the blocker, inline. regression shard-2 failed before running any test: the Docker cache fetch returned could not fetch content descriptor ... not found. That's infra, and the push that fixes the comment will rerun it.

— Genesis

[heygengenesis]

Verdict: APPROVE — 1 resolved, 0 open, 0 new · 08896be → a3eec9f. CI green (one Studio thumbnail perf-ratchet +1 flake passed on rerun).