Repository navigation
Description
[webOS 10.3.1 / O22] 1080p H.264/AAC MP4 hangs indefinitely on Direct Play (abnormal audio timestamps, missing PPS)
TV model
LG webOS TV, O22 chipset, O22_DVB, BOARD_DV_1ST (exact retail model to be confirmed).
webOS version
10.3.1 (ponytail-papikonda)
PlxNative version
- Stable: 0.9.0
- Nightly: 0.10.0-nightly-20261026
Installation method
Homebrew Channel on an LG TV with SSH/root access.
What happened
A specific 1080p H.264/AAC MP4 video from a remote shared Plex library hangs indefinitely when playing at Original quality using Direct Play.
The server is accessible and returns HTTP 200, but no video frames are displayed.
Other videos from the same shared Plex server play correctly, including 1080p H.264 content. Local 4K HEVC/Dolby Vision files also play successfully.
Interestingly, the exact same problematic video plays correctly in the official LG Plex application, even with "Force Direct Play" enabled. The actual playback method has not been independently verified through Plex Dashboard.
Forcing transcoding in PlxNative also resolves the playback issue, but I would prefer to preserve Direct Play.
Media details
- Container: MP4
- Video: H.264, 1920x1080, 25 fps
- Audio: AAC stereo
- Duration: 02:07:25
- File size: approximately 6.1 GB
- Plex Media Server: 1.43.5
- Connection: remote shared library
Steps to reproduce
- Open the affected video in PlxNative.
- Select Original quality / Direct Play.
- Start playback.
- Observe that the player remains stuck on "Loading".
- Try seeking to a different position. Playback still fails.
- Change quality to force transcoding. Playback works.
Relevant logs
The source is accessible:
decision: part=directplay ... -> DIRECT PLAY
ff: open https status=200 clen=6113321584
However, H.264 parameter sets are not found:
ff: param_sets=0 bytes
ff: AU#0 size=5 key=false head=[00 00 00 01 09]
A very large audio timestamp difference is reported:
feed: audio starts 45171690ms after the video fed so far
feed a#1 sz=342 fed=45171690666667
The timeline never advances:
timeline playing t=0s/7645s
The Nightly stderr repeatedly reports:
non-existing PPS 0 referenced
PES packet size mismatch
Packet corrupt (stream = 1)
On-screen playback statistics
- HTTP 200, approximately 138.9 MB received
- Video: 0 fps, queue empty
- Audio queue: approximately 2.42 MB
- Observed buffer reserve: 45,246.7 seconds
- No video frames displayed
The reported buffer reserve closely matches the abnormal audio timestamp difference.
Possible cause
This looks potentially related to H.264 parameter-set extraction, abnormal stream timestamps, or packet handling in the FFmpeg demuxing / native feeding pipeline.
It may be related to #600, although that issue involved a different container and was fixed in #874. This issue is still reproducible in the latest tested Nightly, despite that fix.
Expected behavior
The video should start and play at Original quality using Direct Play, without video re-encoding, when the source is compatible with the TV decoder.
Additional information
I have collected event logs from both builds, as well as the Nightly stderr output and playback statistics.
Sanitized log excerpts are available. I can provide additional diagnostics if needed.
Thanks for the project and for looking into this!
Activity
Playback: an MPEG-TS stream under a .mp4 name direct-plays, with its …