What this workflow does

Sometimes you have an image, that needs to be a little bigger to maintain 300 DPI when printed. This workflow allows you expand the area of an image without writing a prompt. It also allows you to use the MaskEditor at the same time to in-paint parts of the image.

How to use this workflow

You need a clip_vision model, inpainting model and a base model capable of the types of images you want to expand. For SDXL you can use:

https://huggingface.co/comfyanonymous/clip_vision_g/blob/main/clip_vision_g.safetensors

https://huggingface.co/diffusers/stable-diffusion-xl-1.0-inpainting-0.1/resolve/main/unet/diffusion_pytorch_model.fp16.safetensors

You will need to set by how much you want to expand the canvas and on which sides. If you use inpainting at the same time the X&Y in MaskComposite needs to match the left and top in Pad image for outpainting.

Tips about this workflow

The clip vision is optional and can be bypassed by joining up the rerouter nodes. Write a prompt in the text_g box instead.

The in-painting is optional and can be bypassed by joining up the rerouter nodes. TO access the mask editor, right click on the load image node and pick Open in MaskEditor. Use the tool to colour over all the things you want to be repainted.

The outpainting may not work if your image is out of sample for what the clip_vision model was trained on.