Skip to content

[AV1] Restore full transforms for target usage 1 - #387

Draft
fjdoria76 wants to merge 1 commit into
intel:mainfrom
fjdoria76:fjdoria76-qsv-av1-quality-low-latency
Draft

[AV1] Restore full transforms for target usage 1#387
fjdoria76 wants to merge 1 commit into
intel:mainfrom
fjdoria76:fjdoria76-qsv-av1-quality-low-latency

Conversation

@fjdoria76

Copy link
Copy Markdown

QSV AV1 quality regressed for quality-focused low-latency encodes because the default AV1 frame header path always selected the reduced transform set, including Target Usage 1.

This restores the intended default: Target Usage 1 uses the full transform set when supported, while Target Usages 2-7 continue using the reduced set. Explicit ReducedTxSetUsed overrides and unsupported-hardware handling remain unchanged.

Validation:

  • git diff --check
  • CMake configuration succeeded
  • Full build attempted but could not complete because the environment is missing va/va_version.h

Fixes: #375

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
@fjdoria76
fjdoria76 marked this pull request as draft September 11, 2026 08:04
@fjdoria76
fjdoria76 marked this pull request as ready for review September 11, 2026 08:04
@fjdoria76
fjdoria76 marked this pull request as draft September 11, 2026 08:05
@fjdoria76

Copy link
Copy Markdown
Author

hi

@fjdoria76
fjdoria76 marked this pull request as ready for review September 11, 2026 08:08
@fjdoria76
fjdoria76 marked this pull request as draft September 11, 2026 08:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

QSV AV1 encoding quality poorer than QSV HEVC, with low-latency settings

1 participant