@mdgate/audio

Convert audio to Markdown in TypeScript.

@mdgate/audio recognizes MP3, WAV, M4A, AAC, Ogg, FLAC, and WebM audio, then hands the file to a transcription callback you provide.

It does not ship a speech model. audio() is not registered by all().

Works in Node.js, Cloudflare Workers, Edge runtimes, and browsers.

npm install @mdgate/audio

Why @mdgate/audio

Audio is not a document format. Documents that software can parse are handled deterministically. Speech needs a model.

This package keeps that split:

audio bytes → your transcription callback → Markdown

• Pure TypeScript
• No Python runtime
• No native addons
• No WASM runtime
• Zero third-party runtime dependencies in this package
• Detects common audio types from their contents, not only the filename
• Not included in all()

What it does

The converter sniffs the audio type, then calls your (input) => markdown function with { bytes, mime }.

The whole file is passed through. mdgate does not decode or resample the audio.

Without a callback, conversion fails as unsupported.

Usage

import { create } from '@mdgate/core';
import { audio } from '@mdgate/audio';
import { ai } from '@mdgate/ai';

const media = ai({
  baseURL: 'https://api.example.com/v1',
  apiKey: process.env.API_KEY!,
  model: 'your-model',
});

const read = create([
  audio(media.convertAudio),
]);

const markdown = await read(bytes, {
  path: 'meeting.mp3',
});

Or pass any transcription function:

import { create } from '@mdgate/core';
import { audio } from '@mdgate/audio';

const read = create([
  audio(async (input) => {
    // Send input.bytes to the speech model you already use.
    return '...';
  }),
]);

Node.js

import { readFile } from 'node:fs/promises';
import { create } from '@mdgate/core';
import { audio } from '@mdgate/audio';

const read = create([
  audio(transcribe),
]);

const bytes = new Uint8Array(await readFile('meeting.mp3'));
const markdown = await read(bytes, { path: 'meeting.mp3' });

Cloudflare Workers and Edge runtimes

import { create } from '@mdgate/core';
import { audio } from '@mdgate/audio';

const read = create([
  audio(transcribe),
]);

export default {
  async fetch(request: Request) {
    const bytes = new Uint8Array(await request.arrayBuffer());
    const markdown = await read(bytes);

    return new Response(markdown, {
      headers: {
        'content-type': 'text/markdown; charset=utf-8',
      },
    });
  },
};

Need more than audio?

If your application needs to read many different file types, use the complete converter set, then add audio() on top. all() does not register it.

npm install @mdgate/converters

@mdgate/audio is one of the single-format packages in the open-source mdgate/converters project.

For AI agents, the same converter architecture can be used to extend read_file from text files to real-world document formats.

License

MIT