cross-posted from: https://lemmy.world/post/51997098

I made a little project called YoutubeTranscoder:

https://github.com/cashilysh/youtube-transcoder

The idea is pretty simple. You give it a file and it turns it into an 8K black and white video. You upload that video to YouTube, and later you can download it or even decode it directly from the YouTube stream and get the exact same file back.

I made it as an experiment to see if YouTube's compression would completely destroy the data. Surprisingly, it works.

It uses zfec for error correction and SHA-256 checksums to make sure the data survives YouTube's compression and that the recovered file is actually identical to the original.

It's a single Python file and works on both Linux and Windows.

Important: CPU encoding is currently too slow and doesn't survive YouTube's compression reliably. For now, NVIDIA GPU encoding is the only reliable option I've tested.

You can try it yourself with this test video:

https://www.youtube.com/watch?v=tljK1QxLLA0

The video contains a 40 MB ZIP file with some memes and videos. You can decode it directly from YouTube with:

python VideoTranscoder.py -d https://www.youtube.com/watch?v=tljK1QxLLA0

Full disclosure, the entire project was made with Claude and DeepSeek, since I'm not a professional programmer. 😄

It's obviously not meant to replace proper cloud storage, but I thought it was a pretty fun experiment and could be interesting to data hoarders/file sharers.