Media Downloader
Downloads one audio or video file from a public URL, with a hardened backend
Problem
Fetch one audio or video file from a public media URL through a small hosted service that has to stay safe and inside tight server limits.
Approach
A static GitHub Pages front end calls a Flask and FFmpeg backend on alwaysdata. yt-dlp is the main engine, with a pinned youtube-dl fallback only for direct single YouTube videos; each job runs in an isolated helper process, the generic extractor is disabled and every DNS result must be globally routable, to reduce server-side request forgery risk.
Result
The production profile caps media at 120 minutes and 216 MiB per file, runs one extraction job at a time and allows each client five operations an hour. DRM, paywalled, private and active live media is refused, and ready files are scheduled for removal right after a complete download, or after 10 or 30 minutes by size if abandoned.