It doesn’t work. SynthID is embedded in the image. You don’t need a skill to remove C2PA. There are CLI tools that do that. SynthID and other steganographic watermarks are what matter.
Try generating an image with ChatGPT, remove the watermark, and use SlopOrNot to find out.
nothing against pangram, but you guys do realise you could just hook codex up to the pangram api and have it keep rewriting until the eval says it’s fully human, right?
We don’t make BS farcical claims about 99.98% accuracy. We just build the best AI generated text detector that also happens to run on your iPhone’s Neural Engine.
The latest update to Slop or Not (v1.0.6) adds 100% accurate metadata based detection of AI generated images from the following image generators: AUTOMATIC1111, ComfyUI, Draw Things, Easy Diffusion, Fooocus, InvokeAI, NovelAI and SwarmUI.
Big Leap in Realism
Quality mode enables new levels of visual aesthetics for AI images.
Photorealistic scenes with lifelike lighting, texture, and detail.
Merry Christmas Eve, Carol. Be sure to leave some milk out for Santa.
ALT An illustration of three smiley-face cookies on a plate—one with a bite taken out—on a tiled table, next to a blue carton of milk. In the background, there’s a Christmas tree decorated with lights, and a fire in the fireplace.