AI Video Detector — Check If a Video Is AI-Generated
Upload a clip (MP4, MOV or WEBM, up to 50 MB) to get the likelihood that it was AI-generated, with the ML, OCR and metadata sub-signals behind the score.
Limitations: Results are probabilities, not proof. This is not a dedicated deepfake detector, and real footage can be flagged while AI clips are missed, especially after compression. How results can be wrong →
Upload a video to scan
Video Deepfake Detector
Drop your video here, or browse files
MP4, MOV, or WEBM up to 50MB
How the AI Video Detector Works
Your clip is sent to a third-party AI-detection API. It returns one AI-generation probability plus three sub-signals — an ML model, OCR analysis of text in the clip, and metadata inspection — and we show each of them so you can see what drove the score.
ML Signal
The provider’s machine-learning model estimates whether the visuals look AI-generated. It is one input, not a verdict.
OCR and Metadata Signals
OCR looks at any text rendered in the clip, and metadata inspection looks at the file’s technical details. Either can come back empty when there’s nothing to read, and re-encoded files often lose their original metadata.
Frequently asked questions
MP4, MOV or WEBM clips, up to 50 MB per upload.
It depends on the clip and on the third-party provider's load. Video takes noticeably longer than text or images, so keep the page open until the result appears.
The provider combines an ML model with OCR and metadata checks into one AI-generation probability, and we show each sub-signal. It's a probability, not proof — treat it as one signal alongside what you already know about the clip.
It isn't a dedicated deepfake or face-swap detector. It estimates whether a clip as a whole looks AI-generated, and it returns no separate face-manipulation score for video. For a face-manipulation signal on a still frame, use the AI Image Detector — with the same caveats.
Yes, within the free tier: 3 free scans in total, shared across all five tools (not per day or per month). No signup needed to start.
If you have the file (MP4, MOV or WEBM, under 50 MB) and the right to upload it, yes. Re-encoded or screen-recorded copies lose detail, which makes results less reliable.
Yes. Compression and low resolution hide the fine detail the model looks for, so results on low-quality copies are less reliable in both directions.
The clip is sent to a third-party AI-detection API to produce the result. If you're signed in, your scan history keeps the file name and the result, not the video.