Generative Broadcasting

Design your generative radio station.

The new Radio Starlight lets you compose a show segment‑by‑segment, bind each segment to Feeds, APIs, or Pages, and steer it with your own prompt, personality, and BGM sound set. Publish everything with an open protocol.

Capabilities

Build shows that stay fresh and grounded

Multi‑source content engine

Mix Feeds (RSS, Atom, JSON), connect API content, or scrape any Page. Bind them to segments to keep your show fresh and grounded.

Manual program designer

Lay out segments, pick sources, and craft per‑segment prompts. Swap BGMs between segments for dynamic shows.

Generative AI, your way

Use cloud LLMs like OpenAI API, or run on‑device with Apple's Foundation Models. Generate cover art and avatars.

Open publishing protocol

Publish your designed program, generated broadcasts, personalities, sound sets, and the exact sources you used via the StarlightCatalogService protocol.

Sources & Models

Plug in what you need, per segment

What you can connect

  • Feeds — RSS, Atom, JSON endpoints.
  • API content — Call APIs and bind results to a segment.
  • Page content — Point to any web page, summarize and discuss.

Each segment can use a different source and a different prompt.

Generative AI options

  • Text — Any API and Apple Foundation Models (on‑device) for talk, narration, summaries.
  • Images — Any APIs and Image Playground (on‑device) for program cover images and avatars.

Availability depends on hardware/OS. And you need your own API Key if the service needs.

Open Catalog

Publish with an open catalog

Share your designed program, generated broadcasts, personalities, sound sets, and even the feeds/APIs/pages you used, via a portable, open StarlightCatalogService schema, or just archived file.

This makes your show inspectable, remixable, and discoverable.

Open Source
Specification & reference implementation
View on GitHub
What’s inside

What gets published

  • Designed program (segment structure + prompts)
  • Generated broadcast (episodes)
  • Personality (voice & style)
  • Sound Set (BGM)
  • Sources (Feeds, APIs, Pages)
  • Generative AIs (API connect spec)
  • Catalogs (Published content set)