Do we have a rule about taking existing artwork and running it through img2img on low strength? I haven't seen it here specifically to my knowledge, but I have seen it in other places, and it is a major concern that many people who are against AI art have — and, unlike most fears about AI art, a legitimate one.

I've used img2img before, but I've always been fully transparent about wherever it was the case. Others might not be quite so honest about it.

And if we do have a rule about it, where should the line be drawn? There is some point where the generated image becomes unrecognizable from the original — as a demonstration, here's a GIF showing a still from the 60s Soviet cartoon 'Glass Harmonica' morphing into an AI image of a presenting hippo —but it can be hard to say where that point IS. A real Ship of Theseus type of situation.

lendrimujina said:
Do we have a rule about taking existing artwork and running it through img2img on low strength? I haven't seen it here specifically to my knowledge, but I have seen it in other places, and it is a major concern that many people who are against AI art have — and, unlike most fears about AI art, a legitimate one.

I've used img2img before, but I've always been fully transparent about wherever it was the case. Others might not be quite so honest about it.

And if we do have a rule about it, where should the line be drawn? There is some point where the generated image becomes unrecognizable from the original — as a demonstration, here's a GIF showing a still from the 60s Soviet cartoon 'Glass Harmonica' morphing into an AI image of a presenting hippo —but it can be hard to say where that point IS. A real Ship of Theseus type of situation.

I believe that content like that should be reported as a trace.