Whisper by OpenAI - A neural net for speech recognition

by•
Whisper is an automatic speech recognition (ASR) system trained on 680,000 hours of multilingual and multitask supervised data collected from the web.

Add a comment

Replies

Best
Tried out the model, great 👍 very good accuracy. Looking forward to the word level timestamp feature. It would enable our transcript based video editing at Visla.
Excellent and free. It's easy to use and the results are very good.

I'm curious Is there any smaller version that coming up that

Can run in rasberry pi for translate speech to text and text to talk back that offline no internet need?

Whisper is honestly one of the most impressive AI tools OpenAI has released. The transcription accuracy is insanely good, especially for different accents and noisy audio, and being able to run it locally is a huge plus for privacy. It’s not perfect though — it can occasionally hallucinate text, so I wouldn’t trust it blindly for medical/legal stuff. Setup can also feel a bit technical if you’re not familiar with AI tools or command-line workflows. But overall, it is something i would use on a daily basis.