Just a guess, but using Stable Diffusion's img2img feature. Take the image of the happy merchant, (or a MS paint drawing that resembles the happy merchant in shape) add prompts like "nigger dancing with blonde coalburner at a wedding reception" or something like that, and then generate a bunch of times until you get a decent result.
And if that isn't how they did it, then theoretically that's how you COULD do it.
AngryWhiteKeyboardWarrior 0 points 1.3 years ago
Just a guess, but using Stable Diffusion's img2img feature. Take the image of the happy merchant, (or a MS paint drawing that resembles the happy merchant in shape) add prompts like "nigger dancing with blonde coalburner at a wedding reception" or something like that, and then generate a bunch of times until you get a decent result.
And if that isn't how they did it, then theoretically that's how you COULD do it.