[Issue]: Trickplay image generation failing

[ericswpark]

Please describe your bug

Running the trickplay image generation task causes the attached log output.

Reproduction Steps

1. Enable trickplay image generation on libraries
2. Run the scheduled task
3. Observe logs

Jellyfin Version

10.9.0

if other:

No response

Environment

- OS: Unraid
- Linux Kernel: 6.1.79-Unraid
- Virtualization: Docker
- Clients: Browser
- Browser: Chrome 124.0.6367.202 (Official Build) (64-bit) (cohort: M124 Rollout) 
- FFmpeg Version: ffmpeg version 6.0.1-Jellyfin 
- Playback Method: N/A
- Hardware Acceleration: VAAPI
- GPU Model: AMD Ryzen 5 PRO 4650G with Radeon Graphics
- Plugins: AniDB, AudioDB, jellyscrub, MusicBrainz, OMDb, Open Subtitles, Playback Reporting, Reports, Studio Images, Subtitle Extract, TMDb, TMDb Trailers, Trakt, Webhook
- Reverse Proxy: nginx
- Base URL: none
- Networking: Bridge/NAT
- Storage: local

Jellyfin logs

[2024-05-12 18:27:00.886 +09:00] [INF] [41] Jellyfin.Server.Implementations.Trickplay.TrickplayManager: Creating trickplay files at 320 width, for "/data/media/TVShows/Demo/test.mp4" [ID: 9859214e-8bfc-87c4-a04e-642cf2f6f57d]
[2024-05-12 18:27:00.886 +09:00] [INF] [41] MediaBrowser.MediaEncoding.Encoder.MediaEncoder: Trickplay generation: "/usr/lib/jellyfin-ffmpeg/ffmpeg -loglevel error -init_hw_device drm=dr:/dev/dri/renderD128 -init_hw_device vaapi=va@dr -init_hw_device vulkan=vk@dr -filter_hw_device vk -hwaccel vaapi -hwaccel_output_format vaapi -noautorotate -i file:\"/data/media/Demo/test.mp4\" -noautoscale -an -sn -vf \"fps=0.5,setparams=color_primaries=bt2020:color_trc=smpte2084:colorspace=bt2020nc,hwmap=derive_device=vulkan,format=vulkan,libplacebo=upscaler=none:downscaler=none:w=320:h=146:format=bgra:tonemapping=bt.2390:peak_detect=0:color_primaries=bt709:color_trc=bt709:colorspace=bt709,hwmap=derive_device=vaapi,format=vaapi,scale_vaapi=format=nv12\" -threads 1 -c:v mjpeg_vaapi -qscale:v 4 -f image2 \"/config/cache/temp/8ceea609869e46e681dcaeff4282d38d/%08d.jpg\""
[2024-05-12 18:27:01.667 +09:00] [ERR] [39] Jellyfin.Server.Implementations.Trickplay.TrickplayManager: Error creating trickplay images.
System.ArgumentException: Can't create trickplay from 0 images.
   at Jellyfin.Server.Implementations.Trickplay.TrickplayManager.CreateTiles(List`1 images, Int32 width, TrickplayOptions options, String outputDir)
   at Jellyfin.Server.Implementations.Trickplay.TrickplayManager.RefreshTrickplayDataInternal(Video video, Boolean replace, Int32 width, TrickplayOptions options, CancellationToken cancellationToken)

FFmpeg logs

No response

Please attach any browser or client logs here

No response

Please attach any screenshots here

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[gnattu]

Try disable hardware encoding for trickplay and see if it would success.

[ericswpark]

@gnattu thanks, seems like the following option was the culprit:

Unchecked it and ran the task again and now it's progressing fine.

I am using VAAPI though; maybe an incompatibility with AMD GPUs? Feel free to close the issue if this is intended behavior.

[gnattu]

The mjpeg encoder has a very limited hardware platform support, and I don't think it is compatible with anything but Intel atm.

[gnattu]

Closing as hardware compatibility issue

[mcronce]

@gnattu I'm seeing this using an Intel GPU (A380) with 10.9.1 when hardware encoding is enabled - for whatever reason, though, I'm seeing ffmpeg's stderr in logs. The error I'm seeing is [mjpeg_vaapi @ 0x557871ed2480] Invalid quality value 472 (must be 1-100). Is there anything actionable that could be done (on Jellyfin's side) to fix this or is it just a matter of needing to leave hardware encoding off until the hardware encoder matures?

FWIW, execing into the container and manually running the ffmpeg command with lower qscale values does yield changes, but the only value that works is 0, which isn't even in the documented valid range for that value (1-31). 1 still yields Invalid quality value 118 (must be 1-100)

[gnattu]

@gnattu I'm seeing this using an Intel GPU (A380) with 10.9.1 when hardware encoding is enabled - for whatever reason, though, I'm seeing ffmpeg's stderr in logs. The error I'm seeing is [mjpeg_vaapi @ 0x557871ed2480] Invalid quality value 472 (must be 1-100). Is there anything actionable that could be done (on Jellyfin's side) to fix this or is it just a matter of needing to leave hardware encoding off until the hardware encoder matures?

FWIW, execing into the container and manually running the ffmpeg command with lower qscale values does yield changes, but the only value that works is 0, which isn't even in the documented valid range for that value (1-31). 1 still yields Invalid quality value 118 (must be 1-100)

Hardware encoders use different definition of quality. For VAAPI, what you are specifying is the quant and will be multiplied by 118 to get the real quality but that one ranges from 1-100. Such conversion probably should be handled by server otherwise the user can only use the default value.