>>98243582So I decided to give that a try. I went with three pictures: one of gamer schoolgirl irys, one of flare, and the third of a random cafe.
This was done with the prompt of "the characters in photo 1 and photo 2 are talking to each other. this takes place in the location in photo 3":
https://files.catbox.moe/cbed53.pngI could correct this further, but I didn't like it.
I decided to be more specific. "Irys is in picture 1, Flare is in picture 2, a cafe is in three. Generate an image of Irys and Flare in the cafe of picture 3." This was the result:
https://files.catbox.moe/hu67un.pngMuch closer, but just the eyes alone didn't fit. The picture uploaded to the left, aka the finalized involved editing the second picture using inpainting from NovelAI. I could do the same on the first, but I find that the second was closer to what I had in mind.
From the looks of it, chatgpt is still dumb and requires further guidance.
Also novelai is still the king at inpainting 