SUB/WAVE

SUB/WAVE is a self-hosted internet radio station for people who want to broadcast music from their own library. An AI DJ picks tracks and speaks between them while every listener hears the same live Icecast stream.

• Listener requests: Listeners can ask for a song, artist or mood in plain language. The DJ finds a match in your library and adds it to the shared queue, but listeners can't skip tracks.
• Scheduled shows: Give shows their own DJ persona, voice and music mood, or base them on a Navidrome playlist. Guest co-hosts and segments for weather, news or custom commentary can fill the gaps between songs.
• Playlist builder: Describe a set or choose a seed track, and the builder selects a running order from your music. The station can also use library tags such as genre and mood when picking tracks.
• Ways to listen: The web player shows upcoming tracks and a live log of the DJ's speech. Listeners can also use native apps for iOS, Android, macOS, Windows and Linux, plus Sonos and internet radios.
• Model and voice choices: The DJ can use local Ollama or an OpenAI-compatible server such as llama.cpp, vLLM or LM Studio. Piper and Kokoro provide local speech; Chatterbox and PocketTTS support local voice cloning through an optional heavy speech component that isn't started by default.

The free, open source software is MIT-licensed and runs on a Linux machine with Docker. Music comes from Navidrome or another Subsonic-compatible server, including Airsonic, Gonic and Funkwhale. Local model and speech options don't require a cloud AI provider. Anthropic and OpenAI are optional cloud LLM choices, while OpenAI and ElevenLabs offer cloud speech.