Autotranscribe

You upload an audio file and get a cached transcription.

This is a microservice approach built on the Cloudflare Stack that shares uploaded audio and cached transcriptions. It uses a WorkerEntryPoint for RPC to connected service bindings.

Usage

This service exposes the following methods:

• listUploads() - List available transcribed uploads
• get(key) - Retrieve a specific upload URL and metadata
• delete(key) - Delete an existing transcription

After it is deployed, you can use it in an existing Worker like so:

[[services]]
binding = "UPLOADER"
service = "autotranscriber"
entrypoint = "Uploader"

const uploads = await env.UPLOADER.listUploads();

Transcripts are auto generated by adding to the uploads bucket created below in the Setup step.

Setup

We are going to provision an R2 bucket, a Queue, and a KV Namespace to cache our results from the Whisper model hosted on Workers AI.

# Create a new object storage R2 bucket named "uploads"
npx wrangler r2 bucket create uploads

# This will enable the bucket to serve the file
npx wrangler r2 bucket dev-url enable uploads
# ✨ Public access enabled at 'https://pub-your-url-here.r2.dev'.

Modify your wrangler.json with your bucket id.

NOTE: You can edit your CORS Headers if you want to access this from another application

# Create a new Queue named "uploaded"
npx wrangler queues create uploaded

# Get notified of creations and deletions to "uploads" bucket via the "uploaded" Queue
npx wrangler r2 bucket notification create uploads --event-type object-create --event-type object-delete --queue uploaded

# Create a KV store where you can cache your results
npx wrangler kv namespace create uploaded-metadata

Develop

npm run dev

If you want Type completion in your worker:

interface Env {
	UPLOADER: Fetcher<import("../autotranscriber/src/").Uploader>;
}

Deploy

npm run deploy