This is an experimental edit lora for Krea 2 to control poses, image crops and camera positions. Still has a lot of training to go to be great, but it's already showing great potential. I've only tested this on DAF-K2T_v1, so do your own testing with other models. I have not tested with character loras yet, curious to see what others are able to do.

How to use in ComfyUI:

Load in a reference image of your character and route it to the TextEncodeQwenImageEditPlus node, just one image (full body seems to work best), but it will work with close-ups as well.

"Zoom in on" is what the lora's trained on so far, but Krea 2 naturally has some capabilities that seem to allow other instructions ("turn her around", "show in profile", etc). The idea was to teach it to not change the pose of the reference and just zoom and enhance a focus area to help with character sheets, character dataset creation, etc. But it's not quite there yet, it does however work very well with maintaining the style, color pallette and background features.

The following are prompts that have shown to work so far:
"Zoom in on her face"
"Zoom in on her body"
"Zoom in on her breasts"
"Zoom in on her pussy"
"Zoom in on her face and bust"
^^^ only these were trained on v1, the following is just Krea 2 being awesome and adapting

"show her from the front, full body from head to toe, standing"
"show her facing away, full body from head to toe, standing"
"turn her around and zoom in on her back" (ass also works well)

I'm currently working on the next version, to try teaching the model to zoom out, from a close up to a full body image. Fingers crossed.

The workflow below is provided in the downloads