whisper-cpp can do about 6x-10x realtime with the older Whisper model on my 2021 M1 Macbook. I use this to transcribe multiple-hour-long podcasts. The tiny mode

whisper-cpp can do about 6x-10x realtime with the older Whisper model on my 2021 M1 Macbook. I use this to transcribe multiple-hour-long podcasts. The tiny model can easily do 30x realtime on this hardware. Here's a thread about running realtime on raspberry pi devices (some tweaks required): https://github.com/ggerganov/whisper.cpp/discussions/166 https://github.com/ggerganov/whisper.cpp/discussions/166