You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Feature Request: Filter downloads by available audio language track (Multi-audio / Auto-dub) #957
Currently, many YouTube channels are uploading videos with auto-translated or multi-language audio tracks (e.g. Spanish dubs).
It would be great to have a native option in Youtarr per channel to only download videos that contain a specific audio language track (e.g., Spanish/es), skipping videos from channels that do not have that language option available yet.
yt-dlp doesn't make this easy to filter on. A video's "language" is only its original audio, and the dubbed tracks are buried in the list of download formats. If Youtarr asked for a Spanish track and the video didn't have one, yt-dlp would treat that as a failed download, not a skipped video. Youtarr would then retry it every run and send failure notifications. Working around that would take a lot of extra code and an extra YouTube request for every video.
The only way that I know to do this is to download every video and use FFmpeg to filter through the audio. This can be done in a Tdarr flow, but doing this is a lot of extra work, but it can be done.
Activity
dialmaster commentedon Oct 26, 2026
I looked into this, and I don't plan to add it.
yt-dlp doesn't make this easy to filter on. A video's "language" is only its original audio, and the dubbed tracks are buried in the list of download formats. If Youtarr asked for a Spanish track and the video didn't have one, yt-dlp would treat that as a failed download, not a skipped video. Youtarr would then retry it every run and send failure notifications. Working around that would take a lot of extra code and an extra YouTube request for every video.
BigFlubba commentedon Oct 27, 2026
The only way that I know to do this is to download every video and use FFmpeg to filter through the audio. This can be done in a Tdarr flow, but doing this is a lot of extra work, but it can be done.