This is a test / proof-of-concept node application for AWS' streaming transcription service.
Currently the mic-to-trancriber integration has a bug that I don't know how to fix. I believe it's somewhere in the transformation of the audio data provided by the mic stream to the binary data expected by the transcription service.
yarn start speaker
This pipes the audio data from a readable microphone stream to a writable speaker stream. This works and proves that the provided microphone stream is working as expected.
Note: Since this writes your microphone to your speakers, you're going to get a feedback loop. Be prepared to quit this when you run it.
yarn start transcriber
This pipes the audio data from a readable microphone stream to a writable transcriber stream, backed by a WebSocket connection to AWS' streaming transcription service. It's heavily based off of AWS' in-browser, mic-transcriber example.
This doesn't work and I'm not sure why. It is clear, however, that the issue is somewhere in passing the data from the mic to the transcriber and not with sending or receiving data from the transcription service. We know this because the WebSocket client comes up correctly and receives responses from the service when requests are made.
You can see this by running start with the transcriber command. You'll see periodic log lines that include the responses from the service. Since the audio data it's receiving is bad, the responses have no results like the following
Transcript: {
"Results": []
}
Get the bug fixed. I've reached out for help by
- creating an issue in the original amazon-transcribe-websocket-static example
- asking a question on stackoverflow