This is the SDXL version of the previously released FLUX Image Refiner. It detects and masks any part of an image using Florence2, then redraws it using SDXL. You need to configure any checkpoints or LoRA

TIME COST

For Free Generation users, here are the estimated time costs.

The time it takes to generate is approximately 20 seconds to 40 seconds.

HOW TO USE

Upload your image, enter the target area for editing in the “TARGET” field, and input your edit details in the “EDIT” field.

Next, set the denoise value. The default is 0.80. Higher denoise values result in stronger corrections.

It produces good results for body corrections, but it doesn't seem suitable for correcting fingers or toes. If you want to correct fingers, use FLUX REFINER or HAND REFINER.

BaseModel/LoRA/VAE

You can use this without changing settings, but if needed, you can change the base model or VAE, or add LoRA.

Additionally, some SDXL base models do not appear to support Inpaint, and images may not render properly.

sdxl REFINER (sdxl + florence2 )

Notes & Commentaires