Hey Product Hunt! 👋 We built FrameGuard after working with a platform that had a serious re-upload problem — users constantly re-encoding, cropping, and mirroring videos to dodge detection. pHash failed immediately on anything real. We ended up building on CLIP visual embeddings + CNN14 audio fingerprinting, with a lot of painful iteration on the scoring formula — about 20 versions before we hit numbers we were happy with. A few things that surprised us along the way: — Audio-only matching caused false positive explosion (trending background music matches unrelated videos). Audio is now corroborating-only. — A chain ratio penalty we thought was clever caused 58.9% recall regression. Removed entirely. — Structural score softening introduced 15 extra false positives per run. Also removed. Current benchmark: F1 0.995, 99.7% precision, 99.4% recall. Free tier is available — no credit card required. Happy to answer any questions about how the detection works.

FrameGuard

Detect re-uploaded videos via API

Detect re-uploaded videos via API