Chirp

Convert the text content of URLs into a podcast feed, each article becoming an episode read by OpenAI's TTS API.

🚧 Chirp is in early beta state – it works, but expect quite a few rough edges.

Getting Started

You'll need an OpenAI API key: https://platform.openai.com/api-keys. I also recommend to set a monthly budget: https://platform.openai.com/account/limits.

Quick Start

To give Chirp a quick spin, use:

docker run -d \
  --name chirp \
  -v /home/your-chirp-data:/data \
  -e OPENAI_API_KEY=your-openai-api-key \
  -p 3000:3000 \
  ghcr.io/alxjpzmn/chirp:latest

Setup

For a more persistent setup of Chirp, it's best to have a Redis instance running separately, e.g. through their official Docker image. If you do so, don't forget to set the two Redis-related environment variables (more on environment variables below).

You can configure the following environment variables (if you run Chirp via bun run dev, just create a .env file inside the packages/core directory and add them there):

Name	Effect
OPENAI_API_KEY	Mandatory, authenticates you against OpenAI's TTS API
PASSWORD	Optional, adds basic authentication using JWTs for your instance. Beware that your feed itself and its data are always public since podcast players don't play well with authentication.
MAX_ARTICLE_CHARS	Optional, this is useful for development if you want to test things and don't want to rack up bills with OpenAI. I use 200 in development.
SSL	Optional, in case you have Chirp served via HTTPS you will need to set this to true or the feed won't work. Defaults to false.
PORT	Optional, defaults to 3000.
DATA_DIR	Optional, defaults to the directory Chirp is being executed from.
REDIS_HOST	Optional, defaults to localhost. If not specified, a Redis instance will be spun up inside the Chirp container.
REDIS_PORT	Optional, defaults to 6379.

Notes

• Beware that Overcast and Castro seem to expect a publicly available podcast feed. Apple Podcasts and direct RSS also work within a e.g. a VPN network.
• Getting the TTS file (and merging it via FFmpeg ) takes some time. For longer texts, expect a couple of minutes. I'll be adding more verbose progress indicators.
• If you set a server password, this will currently only put the /api REST endpoints behind a JWT validation. The podcast feed and its required files (e.g. the feed cover) plus the websocket endpoints won't be requiring a valid token.

Developer Guide

• PRs are more than welcome! If you'd like to make improvements to the code, add a new feature, fix a bug etc., feel free to just work on a separate branch and open a PR, I'll review merge to main asap.
• To contribute, you'll need the Bun runtime: https://bun.sh.
• Chirp is organized as a monorepo, with a web client using React with Vite and the core (backend) using Bun. There's a shared package to share types and utility functions between the two.
• In case you've never spun up Redis before, but pulled the Docker image already, you can run bun run redis inside the main directory.
• To spin up both the Vite development server and the Bun backend, just run bun run dev inside the main directory.

Acknowledgements

Chirp depends on a lot of other projects. Here's an (incomplete) list of them:

• OpenAI TTS API: I've tried other TTS APIs, but this one has the best quality so far in my view.
• Bun: hyped for a reason, it really is that fast, well documented and convenient to use
• Elysia: effortless APIs with Bun
• Chakra UI: It's ridiculously simple to hack together a decent looking web app with it
• Phosphor Icons: some of the most beautiful and versatile icons around
• Readability: extracts the text content of an URL, this is what Firefox uses for its reader view and the folks at the Mozilla foundation were kind enough to pack it into a Node lib, too
• FFmpeg: https://www.youtube.com/watch?v=9kaIXkImCAM