Diffusers
Safetensors
English

Object-removal feature not working

#1
by makisekurisu-jp - opened

The model card states “support inpaint (image-fill), outpaint (image-extend), eraser (object-removal),” but in my testing it can’t perform object removal. Could you tell me what might be going wrong?

Prompt: remove fox ears, "onereward" written on the whiteboard.
ComfyUI_temp_zxmyp_00002_.png

bytedance-research org

The base model flux fill have undergone heavy SFT for object generation, so the improvement on removal is not obvious. We will release a lora for object removal separately. And it will be better
BTW, the prompt for object removal should be "remove", not extra text append.

bytedance-research org

lora updated , see the end of README

bytedance-research org

Note that it is not an instruction-based model like FLUX kontext, it requires descriptive text.

lora updated , see the end of README

Nice!!

ComfyUI_temp_npjnt_00001_.png

makisekurisu-jp changed discussion status to closed

Sign up or log in to comment