Deepfake Explainer

Upload an image to see the classifier reading and the signals behind it. It provides both the evidence that supports the decision and the evidence that weakens it, as well as how much each different piece of evidence contributed. This process is based on image samples, so it will miss data in regions it does not sample. Use the outputs as diagnostic evidence, not authoritative answers.

Much of this data is highly technical: hover or click on the ? icons for brief explanations.