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
One common failure mode I've found for whisper is that technical terms or proper names will often get mangled. It would be great to have a simple way to edit all instances of "similar" tokens at once.
For example, I transcribed Lilith Wittmann's cccamp23 talk and whisper (medium model) would often recognize "Bonify" (a company name) as "Bonifai", or "Schufa" (another company name) as "Schufer".
It would be great to have an easy way to modify all instances of the same or very similar tokens at once.
Alternatively, a "quick manual check" mode that would allow me to correct all low-confidence words in a single pass would be helpful. I'll file another issue for that.
The text was updated successfully, but these errors were encountered:
One common failure mode I've found for whisper is that technical terms or proper names will often get mangled. It would be great to have a simple way to edit all instances of "similar" tokens at once.
For example, I transcribed Lilith Wittmann's cccamp23 talk and whisper (medium model) would often recognize "Bonify" (a company name) as "Bonifai", or "Schufa" (another company name) as "Schufer".
It would be great to have an easy way to modify all instances of the same or very similar tokens at once.
Alternatively, a "quick manual check" mode that would allow me to correct all low-confidence words in a single pass would be helpful. I'll file another issue for that.
The text was updated successfully, but these errors were encountered: