Listen is a small dictation app built around the part of language learning that is annoyingly hard to fake: understanding what somebody actually said at normal speed.
You get a short clip from a real YouTube video, listen as many times as you need, and either type the whole sentence or fill in a few missing words. Listen checks your answer locally, shows exactly what changed, and moves straight on to the next clip.
If you already have Docker, this is the shortest way to a working development setup:
docker compose up --buildThat starts PostgreSQL, pushes the schema, and runs the Next.js development server with hot reload at http://localhost:3000. Source files stay mounted from the host; node_modules, the Next.js cache, and database data live in named volumes so Linux dependencies do not leak into your local checkout.
If the database is empty, just open the app. Listen will find its first clips automatically and start practicing as soon as they are ready
There’s no final lesson or finish line, once you finish one clip, another is ready to go :) The indexer can keep pulling in newly published, captioned videos too, so the practice pool keeps growing with you!!
That's it, enojy :)