Crowbarr-CUDA ghcr.io/amanofvaly/crowbarr:latest-cuda https://github.com/amanofvaly/crowbarr/pkgs/container/crowbarr bridge bash false https://github.com/amanofvaly/crowbarr/issues https://github.com/amanofvaly/crowbarr https://github.com/amanofvaly/crowbarr/blob/main/docs/unraid.md Audio-grounded subtitle synchronization and transcription for Sonarr, Radarr and Bazarr. Crowbarr watches your Sonarr and Radarr libraries and keeps subtitles in sync with the dialogue. It waits for Bazarr to find a subtitle, checks its timing against the audio with Whisper, and corrects it only when the audio says it is wrong. When no subtitle turns up, it transcribes one. Results are written as a new .crowbarr.en.srt file next to the video. Existing media and subtitles are never modified. This template uses an NVIDIA GPU for speech recognition. Install the Nvidia Driver plugin first, then select CUDA under Settings after the first start. Without a GPU, install Crowbarr instead. Setup: map /tv and /movies to the same shares Sonarr and Radarr use. The container runs as nobody:users (99:100), so those shares must be writable by that account. Open the WebUI, create your account, then add Sonarr, Radarr and Plex under Settings. Speech models download into appdata on first use. MediaApp:Video http://[IP]:[PORT:8449] https://raw.githubusercontent.com/amanofvaly/unraid-templates/main/templates/crowbarr-cuda.xml https://raw.githubusercontent.com/amanofvaly/unraid-templates/main/icon.png nvidia gpu subtitles subsync whisper sonarr radarr bazarr srt Nvidia Driver plugin and a CUDA-capable NVIDIA GPU. Sonarr or Radarr. Outbound internet access on first run to download speech models. --runtime=nvidia --user 99:100 --init --stop-timeout 45 --security-opt no-new-privileges:true --cap-drop ALL MIT 6.10 2026-09-14 ### 2026-09-14 - Initial template.