Repository navigation
feat: encode H.264 and H.265 temporal layers with VA-API - #381
Merged
Merged
Conversation
The VA-API AV1 encoder now takes L1T2 and L1T3, with WebRTC's reference structure and a rate control for each layer, rather than leaving them to libaom. A frame it cannot encode now falls back to libaom.
The VA-API VP9 encoder takes L1T2 and L1T3 too, which Intel GPUs encode, in error resilient mode and with the VP9 information of flexible mode, as libvpx does. The AV1 and VP9 encoders share the layer handling.
The VA-API H.264 and H.265 encoders take L1T2 and L1T3 too. H.265 carries the temporal id in its NAL units and keeps only lower layer pictures in its reference picture sets; H.264 leaves the top layer unreferenced, allows gaps in frame_num and reorders the reference list. Each layer gets its own rate control only where the driver has one that works. The H.264 encoder now uses the shared VaapiEncodeSession.
SendableMetatype
added a commit
to EduGeyser/webrtc-java
that referenced
this pull request
Oct 18, 2026
Brings in the work upstream added from October 15 to 17: temporal layers in the VA-API encoders and VP9 spatial layers in the VA-API decoder (devopvoid#379 to devopvoid#381, devopvoid#383, devopvoid#388); a standalone voice activity detector, and checks against using a disposed one (devopvoid#394); the removal of Logging.logThreads (devopvoid#395); upstream's 0.20.0 release and its release workflow changes (devopvoid#391 to devopvoid#393); and docs (devopvoid#378, devopvoid#382, devopvoid#384 to devopvoid#386, devopvoid#389, devopvoid#390). Fork-side adaptations: - Logging keeps this fork's design, in which thread IDs and timestamps come from system properties read at load time. devopvoid#395's thread IDs that are always on and its runtime Logging.logTimestamps are left out. Its logging test change, which logs at INFO, is taken. - The common WebRTC patches of devopvoid#394 are applied on every platform, as upstream does. The data channels desktop capture patch joins the patch list before it is recorded, so the build record covers it too. - The data channels variant leaves out the standalone voice activity detector, and its test carries the media tag. - The release workflow stays this fork's Gradle one. Upstream's Maven-based split into preparation and tag-scoped publication (release-publish.yml and the publish action) is left out, and so is its release notes link to jrtc.dev, which documents upstream's packages. - The Maven files stay deleted, and the README and the docs' version stay this fork's own.
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
1 participant
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.
The VA-API H.264 and H.265 encoders take L1T2 and L1T3 too. H.265 carries the temporal id in its NAL units and keeps only lower layer pictures in its reference picture sets; H.264 leaves the top layer unreferenced, allows gaps in frame_num and reorders the reference list. Each layer gets its own rate control only where the driver has one that works (not H.264: Intel has none, AMD overshoots with it). The H.264 encoder now uses the shared VaapiEncodeSession.
Builds on #380, which this branch contains; merge it after #379 and #380.