dlnator

a self-hosted tool to pull video/audio from youtube (and anything yt-dlp supports), store it locally, and watch or listen to it through a simple web ui.

what it does

• paste a url, it fetches metadata (title, uploader, duration, thumbnail)
• download as mp4, mp3, or m4a
• stream downloads back through the browser — no re-download needed
• track everything in a dashboard: total jobs, in-progress, completed
• browse finished downloads in a library grid

requirements

• docker + docker compose

running it

docker compose up -d

app comes up on http://localhost:3000 (or whatever port you set in docker-compose.yml). data — jobs.json and all downloaded media — is stored in ./data, which persists across restarts and rebuilds.

by default the container binds to all interfaces, so it's also reachable from other devices on your network at http://<your-machine-ip>:3000 (useful for phones/tablets). there's no authentication, so only do this on a network you trust.

usage

1. dashboard — paste a url under "quick add" to create a job
2. downloads — pick a format (mp4 / mp3 / m4a) to start downloading; download multiple formats of the same source if you want
3. watch — once a format finishes, open it to stream in-browser
4. library — grid view of everything fully downloaded

notes

• supports any source yt-dlp supports, not just youtube
• re-adding the same url won't create a duplicate job — it reuses the existing one
• deleting a job removes its files from disk, not just the entry
• jobs are currently stored in a flat jobs.json file — fine for personal use, not built for heavy concurrent use yet (see planned)

planned

• migrate job storage from jobs.json to a proper database (better concurrency, no race conditions on simultaneous writes)
• search/filter on downloads and library pages
• playlist and batch url support
• subtitle download
• working settings (default format, rescan library, clear failed jobs)
• job retry on failure
• authentication

contributing

this is an early, personal-use tool and there's plenty of room to improve it. issues and pull requests are welcome — whether it's a bug fix, a planned feature above, or something not listed here.

a note on how this was built

this project is hand-designed and hand-reviewed, not vibe-coded — every change was deliberate and checked, not just accepted from a prompt. ai assistance was used along the way (mainly for code review, cleanup, and drafting), but the architecture, decisions, and final code are owned and understood, not blindly generated.