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
Unfortunately the whisper model has no built-in method to do this currently. What you could do though is to extract the video with each audio track separately into multiple files using ffmpeg, then run the CLI to batch process them all, or run them manually one by one using the webui.
You can extract the video and the first audio track for example using this command:
I have a video with multiple Audio Track (like french english ...etc) How can i transcribe a spesific language from all of these languages?
The text was updated successfully, but these errors were encountered: