does nt work at all not even works for single image
I gave a detailed prompt: "Repose the person sitting relaxed on a park bench with legs crossed, casual smile, change background to a cozy autumn forest with falling leaves, preserve facial identity perfectly, identical face features, do not change outfit or hairstyle, realistic lighting matching the scene" : all i get is only one change with autumn leaves and thats about it: too bad
Hi. Thanks for your feedback! As shown in the image above, everything is running. This is the same workflow, and I've set up a scenario similar to what you described.
I gave a detailed prompt: "Repose the person sitting relaxed on a park bench with legs crossed, casual smile, change background to a cozy autumn forest with falling leaves, preserve facial identity perfectly, identical face features, do not change outfit or hairstyle, realistic lighting matching the scene" : all i get is only one change with autumn leaves and thats about it: too bad
one tip is to describe the image you want not "asking"
I gave a detailed prompt: "Repose the person sitting relaxed on a park bench with legs crossed, casual smile, change background to a cozy autumn forest with falling leaves, preserve facial identity perfectly, identical face features, do not change outfit or hairstyle, realistic lighting matching the scene" : all i get is only one change with autumn leaves and thats about it: too bad
one tip is to describe the image you want not "asking"
So that's not supposed to be the case with Flux.2, which is trained specifically to follow instructions. "Describing the image you want" is likely to induce side effects. This maybe works better with the base model, but there's no KV version of the base model yet