WhisperD: linux voice-to-text using OpenAI whisper-1 transcription
5 points by icholy
5 points by icholy
I wrote this as an exercise to learn how to use ioctl & input devices, but I like how it turned out! It does have a hard dependency on pipewire though.
From the title, I thought this was using whisper with whisper.cpp.
But I just misunderstood, it looks this is just streaming the microphone to OpenAI servers. This doesn’t diminish the work of gluing everything together.
That could be a cool next step though, to make it local only since whisper.cpp has some go bindings: https://github.com/ggml-org/whisper.cpp/tree/master/bindings/go
It’s the same whisper model. Switching it to use a local model wouldn’t be difficult, but I don’t have hardware that can handle that.
edit: I just tested and I can run it locally no problem. Thanks for the tip!
The ‘suggest tags’ button seems to have gone, but this should be using the vibecoding tag, not ai. The description of the ai tag says:
Developing artificial intelligence, machine learning. Tag AI usage only with
vibecoding.