How to generate images for different sizes?

[Hongbin98]

According to inference.py, it seems that all images are fixed to 512*512.

How do I generate images with different sizes via insdiff?

[frank-xwang]

Hi! Unfortunately, we don’t support image generation at different sizes by default. However, there are some packages in ComfyUI that can help create multi-size outputs with additional components. You might want to check out the ComfyUI-InstanceDiffusion repository, along with other modules designed for stable diffusion, to achieve this.

btw, there are some YouTube videos to introduce the pipeline of InstanceDiffusion in ComfyUI, you may also want to take a look to better understand how to make changes.

• ComfyUI AnimateDiff Instance Diffusion workflow: https://www.youtube.com/watch?v=jVyd98CwUW0
• This Diffusion Model Are Insanely Great! Instance Diffusion Create Animation In ComfyUI: https://www.youtube.com/watch?v=i0ZbeUNeLCE