Native whisper.cpp server with OAI-like API is now available $ make server && ./server This is a very convenient way to run an efficient local transcription service locally on any kind of hardware (CPU, GPU (CUDA or Metal) or ANE) thx felrock

Amazing! I had to code it myself because wasn’t able to find anyone to expose it. It uses Rust, but calls WhisperCPP under the hood

I cannot believe this is free! Open source FTW. Thank you. Going to try it out today.

This is M2 Ultra using Metal with Base whisper model. For detailed benchmarks on different hardware and model sizes, checkout the release notes of v1.5.0 in the repo

This is incredible. Just what I needed for project. Thank you 🙏

This is great. Thanks GG. One question though. Does whisper now support speaker identification & labelling? It’d be great to be able to output a formatted & speaker ID’ed text. Unless side processing is supported, this feature may only be possible by modifying Whisper core.

Would it be possible to stream audio to the webserver for realtime transcription?

all we need now is a websocket server for realtime audio transcription

Saludos @cibernicola_es agradecería tu opinión sobre esto.

Saludos bro @JassonTesla tu comentario sobre esto.