mirror of
https://github.com/heygen-com/hyperframes.git
synced 2026-09-14 18:01:20 +08:00
xiaye
671f58507f
Three independent costs in `hyperframes check`, none of which changed what the gate detects: 1. Motion-only grid times took the full audit settle. AUDIT_SEEK_OPTIONS pays an unconditional 120ms paint-flush sleep plus a glyph-subset font wait on every seek, but a time that only feeds collectMotionFrame needs neither — __hyperframesMotionSample reads computed transforms and opacity, which the ordered double-rAF already guarantees are committed. At 20fps the motion grid is up to 300 samples, so this sleep dominated any run carrying an index.motion.json. Times that measure text (layout), photograph the frame (contrast) or read caption/frame geometry keep the full settle. 2. `waitForCompositionSettle` slept a flat 1500ms on every page open, after __renderReady, shader pre-render and document.fonts.ready had all already resolved. Replaced with a poll that holds a 400ms floor, then exits on the first pair of consecutive frames with no font subset in flight, capped at the original 1500ms — worst case unchanged, common case ~1s faster. 3. `collectContrast` always took a second full-page screenshot to build the annotated overview, then the pipeline discarded it unless --snapshots was set. The bare `collectContrast(time)` call shape already meant "no snapshots", so that now skips the shot and its overlay entirely. collectGridSamples' per-time layout and contrast blocks are extracted to keep it under the complexity gate after the seek-profile branch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>