SyntheticInteractive study companion
Can today's detectors tell a real face from a fake?
An interactive companion to a study on how off-the-shelf AI-image detectors hold up on real and AI-generated faces. Pick a detector and a condition and see how it scores.
Query the benchmark
Pick a selection and press Run.
AUC is a 0 to 1 score for how well a detector tells fake images from real ones. Picture showing it one fake photo and one real photo: AUC is how often it rates the fake as the more likely fake. 1.0 means it always gets it right, 0.5 is a coin flip (no better than guessing), and below 0.5 means it is getting them backwards.
Put it to work
Public model
I'll run it and put it up on the interactive.
Private model
Share it privately. Weights or API Key.
Self service
Test your own model on the public datasets.
Request a generator
Suggest one to add to the interactive.