Reputation: 18657
I'm developing an application that uses Whisper for transcribing audio files. The content is primarily in one language, but the speaker frequently includes phrases or quotes in another language. This mixed-language content is causing issues—either the non-primary language segments are mis-transcribed or omitted entirely.
I'm wondering if there's a way to configure Whisper to better handle these language switches. Should I be looking into adjusting parameters, or perhaps pre-processing the audio to detect and separate the different languages before transcription? Any advice or best practices on accurately transcribing such audio would be appreciated.
Upvotes: 1
Views: 16