August 26, 2025Aug 26 The whisper-asr-webservice doesn't have an official support thread, so posting here. Mods: feel free to relocate to a more appropriate forum.I've deployed the whisper-asr-webservice from the community apps as a docker container. My goal is to have it perform sub creation via Bazarr. I have Bazarr and Whisper talking, and it will generate subs, but only using my CPU.From the documentation, I needed to have the --gpus all extra parameter set. (which it is by default). But with that setting, it's still using the CPU.I've even setup the NVIDIA_VISIBLE_DEVICES and NVIDIA_VISIBLE_DEVICES variables (with the same values that I have set for my other dockers), but still uses CPU.I also added the extra parameter: --runtime=nvidia but still can't get it to use the GPUI'm sort of at a loss on what to try next. Edited August 26, 2025Aug 26 by Xoron
August 26, 2025Aug 26 Author So I've made some progress. I was able to get whisper-asr-webservice to see and attempt to use my GPU, but doesn't seem like it's using the decoded/encoder engine. My original problem was that when you want to use the GPU (--gpus all which is on by default), I had to use the specific repository that calls the compatible docker. onerahmet/openai-whisper-asr-webservice:latest-gpuBut now when Bazarr tries to generate subtitles for videos, it never completes it, and the GPU sits idle, all while using up loads of memory:
November 23, 2025Nov 23 Old thread but I had the same problem and this docker fixed it for me: https://hub.docker.com/r/agoddrie/srtgen.
March 4Mar 4 On 11/23/2025 at 3:36 PM, ARGO666 said:Old thread but I had the same problem and this docker fixed it for me: https://hub.docker.com/r/agoddrie/srtgen.Can you share your docker settings for Unraid? I have it set up but it dies when it starts. I think it's something to do with my GPU settings.
March 4Mar 4 Author I abandoned whisper ASR and went with subgen (https://github.com/McCloudS/subgen). And it's been working fine ever sinceYou just use whisper asr in the bazarr setup, and point it at port 9000 for subgen
March 4Mar 4 I'm not interested in running bazarr, but it looks like it will work with Plex webhooks so I'll give it a shot. Thanks for the recommendation.
March 5Mar 5 Author I'm not interested in running bazarr, but it looks like it will work with Plex webhooks so I'll give it a shot. Thanks for the recommendation.Ok, now you've peaked my interest. What exactly are you using it with Plex for?
March 5Mar 5 2 hours ago, Xoron said:Ok, now you've peaked my interest. What exactly are you using it with Plex for?Nothing crazy. I sometimes download videos from sites like shvideos.net that have no captioning so I want to generate subtitles for them. I don't use the arr stack, I do it manually, so I've been wanting to automate the subtitle process. Putting the videos in the destination folder where Plex will see them and trigger subgen automagically will be nice.
March 5Mar 5 Author That's exactly what bazarr is for. to download (or in this case generate) subtitles on videos. You should seriously look at bazarr to do that for you.
March 27Mar 27 On 3/4/2026 at 5:54 PM, Xoron said:I abandoned whisper ASR and went with subgen (https://github.com/McCloudS/subgen). And it's been working fine ever sinceYou just use whisper asr in the bazarr setup, and point it at port 9000 for subgenI've been having a hard time getting this to work and integrate with Bazarr. Would you mind sharing a screenshot of your settings? obviously censor out anything sensitive like tokens or your IP
March 27Mar 27 Author I've been having a hard time getting this to work and integrate with Bazarr. Would you mind sharing a screenshot of your settings? obviously censor out anything sensitive like tokens or your IPHope this helps.I run my subgen networking through another docker image. So that part might be different for you.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.