Latent Space · AI Builders · November 2026
Discussing what the FLUX Kontext editing model changed, Podell says image generation used to mean re-rolling prompts and hoping; the point was to edit an image step by step while keeping the rest.
For most people, it's very common to generate something, chat with it, change it. Before this, image generation took quite a while, and you needed to recreate the image. You couldn't change the tasks. You could just get the same character. It wouldn't be the same. You
need to mask something that you don't like. So that's what's going on. And then fill in a bit.
I just wanted to get away from the slot machine, essentially. Everyone had to do the slot machine, prompt and prompt and prompt and prompt. And it was very clear that iterative editing was much more effective. My copyconf
UI was that. It just swapped out 10 different versions. Yeah, yeah.
My favorite part, I still slop out. Generate for generate six, redoing something. Just make 10 iterations.
And then, yeah, and eventually that one's good enough.