Any chance you could provide some guidance on how to use this?

[rpelletierrefocus]

I have successfully built the Docker image and spun it up, but I do not understand how to process a file. Any help would be greatly appreciated.

[mhilker]

The container process starts a HTTP server (default on port 8080).

Most of the HTTP plumbing happens inside the server handler: https://github.com/mhilker/transcriber/blob/master/cmd/server/main.go#L55

Currently, it works like this:

• the request method has to be a POST request
• the content-type has to be audio/webm
• the body contains just a blob of your audio file

If everything works correct, you will get back a JSON response with an object containing the hypothesis and the score.

I've used the API like this (be aware this is FLOW flavoured JS, your milage may vary):

// @flow

type Transcription = {
  hypothesis: string,
  score: number,
}

const url = process.env.REACT_APP_API_URL || '';

export default async function (blob: Blob): Promise<Transcription> {
  const response = await fetch(url, {
    method: 'POST',
    body: blob,
    headers: {
      'Content-Type': 'audio/webm',
    },
  });
  const json =  await response.json();
  console.log(json);
  return {
    hypothesis: json.hypothesis || '',
    score: json.score || 0.0,
  };
}

Please let me know if this helps.