wtf is duetto?

Duetto is a self-hosted music player built for two people to listen together, with an AI companion attached to the room. When one person opens the app and shares a room link with the other, play, pause, and seek actions stay in sync between both listeners in real time. Songs can be shared into the room as chat cards, and status updates announce who played, paused, or liked a track. What makes it different from a typical listen-together app is that the AI companion actually processes the audio of the song being played, using a multimodal model you connect yourself, rather than just reading song titles. It builds up a memory of what each song means to the two people over time, based on their conversation about it, and that memory gets included the next time they discuss the same song. Users can long-press a lyric line to quote it in chat or ask questions about that specific line, and the AI can also take actions in the room, such as playing, queueing, or pausing a song, through inline chat commands. The app also integrates with NetEase Cloud Music for search, playlists, and daily recommendations, supports adding your own songs by pasting a direct audio link, and can be installed as a progressive web app with lock-screen playback controls. On the technical side, the frontend is built in React without a build step, and the backend runs on Express with WebSocket for syncing, storing long-term data like song archives and chat memory in SQLite. Running it requires Node 24 or newer, installing dependencies, and starting the server, after which you configure your own AI provider's endpoint and API key inside the app. A PIN-based access gate protects the deployment once set up. This project is aimed at developers comfortable self-hosting a small app for themselves and one other person, such as a partner or close friend, rather than a public multi-user service.