The situation
Text-to-music only feels magical if the wait is short and the first attempt is good. The product had to take a plain-language prompt — 'dreamy lo-fi about late-night Mumbai' — and return a complete, listenable track with vocals, not a loop the user then has to assemble themselves.
Stack
- React
- TypeScript
- Generative audio
- Credits & billing
- PWA
What we did
- 01Prompt-to-track pipeline that composes, performs and produces in one pass, with genre and language selected from the prompt itself.
- 02Streaming-first playback so a track can be auditioned the moment it is ready, rather than after a full download.
- 03Credit system with a free tier that needs no card, so the first song costs a visitor nothing but a sentence.
- 04Community feed of published tracks, turning single-use generations into something worth coming back to.
- 05Installable as a PWA, so mobile listeners get a full-screen app without an app-store round trip.
The outcome
Autunes runs publicly at autunes.com with generation across 70+ genres and 50 languages, a browsable community feed, and a creator marketplace with royalties in build.