The whisper recognizer has been implemented using the stream.wasm example in whisper.cpp. It is fully functional but currently it has the following limitations:
Fixed model (quantized tiny.en)
Fixed thread count (8 or max thread count)
Need to hit mute and unmute to start after a refresh, due to autoplay policy
Naive algorithm that transcribes audio every five seconds
The whisper recognizer has been implemented using the stream.wasm example in whisper.cpp. It is fully functional but currently it has the following limitations: