Hey everyone, I’m facing an issue with Whisper: it’s returning unwanted text in certain cases. For instance:
When the audio file is blank or contains music, it still generates a transcript.
If the mic is left open for a while, it adds random text for that duration.
I’ve already fixed filler utterances and similar issues using prompts, but I need the transcript to reflect exactly what the user speaks (including all fillers and repetitions). I’ve tried different prompts and adjusted settings, but nothing seems to resolve this.
Can anyone guide me on fixing this—be it through prompts, settings, or an alternative platform that provides a pure transcript? Thanks in advance!