Fix playback issues with some h264 videos on native & Safari #8850
+125
−42
Conversation
This file contains 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
|
Web viewer built successfully. If applicable, you should also test it:
Note: This comment is updated whenever you push a commit. |
Wumpf
force-pushed
the
andreas/video-aggressive-enqueueing
branch
from
b830460 to
c6670ba
Compare
Wumpf
commented
Jan 29, 2025
Comment on lines
+277
to
+278
| // (potentially related to:) TODO(#7327, #7595): We don't necessarily have to enqueue full GOPs always. | ||
| // In particularly beyond `requested_gop_idx` this can be overkill. |
There was a problem hiding this comment.
this is a pre-existing issue, it just wasn't referenced here before
Comment on lines
226
to
+228
| // By resetting the current GOP/sample indices, the frame enqueued code below | ||
| // is forced to reset the decoder. | ||
| self.current_gop_idx = usize::MAX; | ||
| self.current_sample_idx = usize::MAX; | ||
| self.last_requested_gop_idx = usize::MAX; |
There was a problem hiding this comment.
this is a bit wonky imho - why not just self.reset directly. A: This would lead to a double reset.
In the interest of keeping the changes here more minimal I'm not touching this piece
emilk
reviewed
Jan 29, 2025
jprochazk
approved these changes
Jan 29, 2025
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Outdated suggestions cannot be applied.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.
Related
ffmpeg-cli& SafariWebCodecdecoder requires additional frames enqueued for some videos in order to produce frames + spurious errors #8848What
Works around the issue outlined in #8848
In a nutshell: For certain videos we would only ever receive frames that are 16 frames behind of what we enqueued, completely disregarding the fact that we always enqueue full GOPs (I repeatedly confirmed that this is the case with different tools). This affects only h264 on the ffmpeg-cli & WebCodec @ Safari decoders. The exact circumstances that make some videos require this and other not are unknown at this point.
This is "solved" here by making our frame enqueing process more flexible and allowing decoder implementations to specify a "minimum look ahead" number of frames.
Note that oftentimes we still enqueue much more frames regardless: We still want to enqueue the entire current & next GOP for smooth transitions. Typically two h264 GOPs encompass a lot more frames than requested here (citation needed). However, the tested problematic videos operate with GOPs that contain only two frames (single I & P frame) so there this actually makes a difference.
Testing: Beyond the actual video, I tested various samples form our internal library of "curious videos to test" to ensure that the adjusted enqueuing logic doesn't regress anything. (personally, I think it's actually much nicer independent of the workaround :))