Hacker News Reader

Will the lack of existence of any stable fingerprint be a valid heuristic for the detection of a deepfake?

Should be straightforward to train a model to add these based on a few examples. Could last for a short bit though.