Testing Whisper with Transformers.js

[shiffman]

Right now this loads from a wav file, it would need to connect to real-time audio and the mic. This may be too difficult to do in node so may want to investigate some other service to do the "speech to text".

[shiffman]

Some leads on how this might work?

https://www.npmjs.com/package/node-record-lpcm16

  audioStream.on('data', async (chunk) => {
    const output = await whisper(chunk);
    console.log('Transcription:', output);
  });