Repository navigation
GPU encoding fails on Intel N100 #36
Copy link
Copy link
Open
Description
I'm unable to take advantage of GPU encoding on my N100. I believe I have everything configured properly, but encoding fails. I'm running in Docker, and I've exposed my GPU to the container. I've also set VERTD_VAAPI_DEVICE_PATH to /dev/dri/renderD128.
Here are the logs from vertd when I try to convert a video:
[2025-12-01T02:41:10Z INFO vertd] starting vertd
[2025-12-01T02:41:10Z INFO vertd] working w/ ffmpeg 6.1.1-3ubuntu5 and ffprobe 6.1.1-3ubuntu5
[2025-12-01T02:41:10Z INFO vertd] using VA-API device path from environment variable VERTD_VAAPI_DEVICE_PATH: /dev/dri/renderD128
[2025-12-01T02:41:10Z INFO vertd] detected an Intel GPU -- if this isn't your vendor, open an issue.
[2025-12-01T02:41:10Z INFO vertd] using VA-API device path: /dev/dri/renderD128
[2025-12-01T02:41:10Z INFO vertd::http] http server listening on 0.0.0.0:24153
[2025-12-01T02:42:05Z INFO vertd::http::services::upload] uploaded file: IMG_3312.mov
[2025-12-01T02:42:12Z INFO vertd::converter] running 'ffmpeg -hide_banner -loglevel error -progress pipe:1 -hwaccel vaapi -vaapi_device /dev/dri/renderD128 -hwaccel_output_format vaapi -i input/87b78cee-7242-4434-88b3-c0bbaee9d2aa.mov -c:v h264_vaapi -c:a aac -strict experimental -preset slow -b:v 26297690 -map_metadata 0 -map_chapters 0 output/87b78cee-7242-4434-88b3-c0bbaee9d2aa.mkv'
[2025-12-01T02:42:12Z ERROR vertd::converter] [h264_vaapi @ 0x55ec7f77a400] No usable encoding profile found.
[2025-12-01T02:42:12Z ERROR vertd::converter] [vost#0:0/h264_vaapi @ 0x55ec7f77a040] Error while opening encoder - maybe incorrect parameters such as bit_rate, rate, width or height.
[2025-12-01T02:42:12Z ERROR vertd::converter] Error while filtering: Function not implemented
[2025-12-01T02:42:12Z ERROR vertd::converter] [out#0/matroska @ 0x55ec7f6b3a00] Nothing was written into output file, because at least one of its streams received no packets.
[2025-12-01T02:42:12Z ERROR vertd::http::services::websocket] job 87b78cee-7242-4434-88b3-c0bbaee9d2aa failed
[2025-12-01T02:42:12Z ERROR vertd::http::services::websocket] failed to handle job failure: environment variable not found
Is there something I'm doing wrong on my side? Any other info I can provide to help?
Metadata
Metadata
Assignees
Labels
No labels
Activity
azurejelly commentedon Dec 1, 2025
Does it work if you run it directly on the host machine?
raddevon commentedon Dec 1, 2025
@azurejelly Yes, in fact, I was able to use this command from within my container, and it seemed to be working:
It appears just the command built by vertd doesn't work for me.
UPDATE: Sorry, I misunderstood your question. I think you're asking if I can run the vertd-generated ffmpeg command on the host. No, I can't. I get the following:
azurejelly commentedon Dec 1, 2025
sorry, i meant whether vertd works when it is run directly on the host machine (i.e. not in a docker container) using one of the binaries from https://github.com/VERT-sh/vertd/releases. i wanted to see if it could be caused by some random missing dependency on the container, though i highly doubt that's the case
raddevon commentedon Dec 1, 2025
Not sure the best way to test on the host, but I did it with this script:
Test script
I'm afraid I did even worse here. The server logs this when I run the script and the upload starts:
JovannMC commentedon Dec 15, 2025
can you try using it through the VERT UI instead of your script and let us know what errors you get there? and does running it on your local machine outside of docker work (as @azurejelly said)?
raddevon commentedon Dec 15, 2025
When I run vertd on the host (no options), I get this:
If I run it with
--gpu intel, it seems to start fine. The video conversion doesn't work though. I get this:Developer tools console output
upmcplanetracker commentedon Sep 6, 2026
I think this has to do with the flags that are hard coded in the container. The repo is trying a one size fits all approach but that doesn't work unfortunately. When you ran it by hand you had different flags and it worked.