Reverse Video Search

Reverse Video Search exists for a specific and increasingly common problem: a clip is circulating, and you need to know where it actually originated before you publish, cite or dispute it. Rather than relying on a single signal, it works multi-modally — visual fingerprinting to find re-encoded and cropped copies, audio fingerprinting plus Whisper-based transcript matching to catch reuploads where the picture has been altered but the sound has not, and OCR or watermark detection to surface channel logos and captions burned into the frame. That combination matters because the standard evasions (mirroring, cropping, re-encoding, adding a border) defeat naive frame matching but rarely defeat all three at once. For creators, it doubles as an enforcement tool: it can assemble DMCA reports for unauthorised reuploads of your own footage. There is also a deepfake detection pass for cases where the question is not who posted it but whether it happened at all. The honest caveats: provenance tooling gives you evidence, not certainty, and no matching system has complete coverage of every platform — a negative result means "not found", not "original". Deepfake detection in particular should be treated as one input among several rather than a verdict. Used the way a journalist or rights holder would use it — as the first pass of an investigation — it is a serious piece of kit.