Repository navigation
Investigate #148: frame rate + NR/Sharpen probe report and regression test - #151
Merged
Merged
Conversation
StuartCameronCode
marked this pull request as ready for review
November 10, 2026 06:34
Issue #148 reported black patches once Noise Reduction and Sharpen were added to a frame rate conversion. It does not reproduce on macOS arm64: across 5 denoisers x 4 sharpeners x FlowFPS/RIFE x 8/10-bit, plus a fast pan and hard cuts, the share of synthesized-frame pixels darker than both real neighbours stays at the 0.001-0.005% noise floor with or without the filters. FlowFPS at 10/16-bit matches 8-bit, and RIFE's out-of-range input comes from the source, not the sharpener. No pass is changed. The new heavy test pins that on all four nightly platforms, which covers what this machine could not: MVTools v24 (Windows, macOS x64) and RIFE on non-Apple Vulkan drivers. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The mean over ~40 synthesized frames moves only ~0.03 points for a black block confined to one frame, under the 0.05 tolerance. Also assert the worst single frame stays within 0.25 points of frame rate alone; note the correction in the report and ENGINEERING_NOTES. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
StuartCameronCode
force-pushed
the
agentbob/6-weird-pipe
branch
from
November 10, 2026 09:15
ae4608b to
c54806c
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Investigates #148: odd frames in frame-rate conversions that were combined with Noise Reduction and Sharpen.
Summary
The bug did not reproduce on macOS arm64, so this PR changes no pipeline code. It adds the evidence and a regression guard:
docs/reports/issue-148-frame-rate-combined/:index.html, captioned images, and the scripts behind every number. It compares FlowFPS and RIFE with the frame-rate pass alone against frame rate plus the default Noise Reduction and Sharpen. The probes run the template's own{{#FRAME_RATE}}block, not a copy.docs/reports/README.md.app/test/integration_frame_rate_combined_test.dart(FlowFPS + RIFE). Its metric is the share of pixels in each synthesized frame that are darker than both of the frame's real neighbours. It asserts two things against frame rate alone:Two candidate fixes were considered and declined, because neither had a measured defect behind it and either would change the output of every frame-rate job:
Testing
flutter analyzeis clean.Review closely
scripts/run.shexportsVAPOURSYNTH_PLUGIN_PATH. It has been inert since R74, so it is harmless but misleading.Follow-up
Ask the reporter on #148 for:
Refs #148