User enters a YouTube URL.
Frontend sends a POST request to /api/download with the target URL.
yt-dlp powered • Local storage • No database
Turn AutoCaptions into AnyDownload.
AnyDownload reframes the project around a broader download workflow: inspect a YouTube URL, return available media options, and let users fetch MP4, MP3, subtitles, captions, transcripts, or thumbnails without introducing a database layer.
API asks yt-dlp for the JSON manifest.
Return title, duration, thumbnail, subtitle languages, and downloadable formats before any heavy transfer begins.
User selects MP4, MP3, subtitle, caption, transcript, or thumbnail.
Frontend can render clear actions such as Download 720p, Download MP3, Get subtitles, or Save thumbnail.
API streams the chosen file back directly.
Prefer download → stream → save. Avoid download → persist → resend when running on a serverless host.
Feature direction
One project, multiple output types.
Instead of staying limited to auto captions, the project now positions YouTube as the main entry point for multiple output types while keeping the experience simple enough for static marketing pages and future API integration.
✅ Open source and ready for incremental backend work
Planned outputs
- MP4 video downloads
- MP3 or extracted audio downloads
- Subtitle and caption track downloads
- Transcript export and thumbnail saving
Technical notes
Designed around Vercel-style limits and local-first storage.
The current concept explicitly calls out serverless constraints: large downloads may exceed execution limits, streaming is safer than saving to disk, and the eventual backend should rely on a bundled yt-dlp strategy rather than assuming Python is available.