I occasionally get a single hallucinated word (more like a mis-transcription) wh... | Hacker News

HN2new | past | comments | ask | show | jobs | submit

		nmfisher on Feb 21, 2023 \| parent \| context \| favorite \| on: Whispers of A.I.’s Modular Future I occasionally get a single hallucinated word (more like a mis-transcription) where the audio contains a clunk/bang/cough/etc, but I've never had full hallucinated phrases from clean silence. There are a couple of GitHub discussions on the Whisper repository with various fixes/hacks to deal with it: https://github.com/openai/whisper/discussions/679 https://github.com/openai/whisper/discussions/813 If you get a chance, I encourage you to try out the other newer models I mentioned, I think you'd be very impressed.

pixl97 on Feb 21, 2023 [–]

I don't see this much different than what commonly happens with humans when we hear our named called when it was some environmental noise.

As for the silence, I wonder why the the model even receives it. I would think a lot of that would be compressed out of existence to save bandwidth.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact