4 ms·
Port of OpenAI's Whisper model in C/C++
- pbowyer 4y agoI'm sharing this because on my M1 Mac I couldn't get Whisper to work with MPS. CPU transcribing was taking around 2x the audio time. With Whisper.cpp with 4 threads and using the base.en model, a 40 minute podcast transcribed in 130 seconds.