This is an extension for AUTOMATIC1111/stable-diffusion-webui which can generate infinite loop videos in minutes.
What it does is repeatedly transformed through outpainting, transitioning from the previous version to the newly created one.

To prepare an initial image, you can either generate one using txt2img or find one on the internet. Just ensure that its aspect ratio is 1:1.

Send the image to img2img/inpaint and ensure that it has the same width and height. You can refer to my parameters listed below.
Prompt is needed, you may use the same prompt as in txt2img if you prefer.

From the Script dropdown menu, select Infinite Zoom and leave it as default if you don't understand what it does.

1. You may want to use inpaint model sd-v1-5-inpainting.ckpt etc.
2. To achieve optimal results, set the Mask Width (mask_width < image_width / 2) to 1/4 of the image width. For example, if your image is 1024x1024 pixels, try setting the mask width to 256 pixels.

You may notice a distinct frame or seam around the image, although it's less noticeable in the original diffusers version.
I'm currently investigating the cause of this issue and would greatly appreciate any assistance you can provide.

I'm just porting it to AUTOMATIC1111.
All credit goes to the original creator https://github.com/BalintKomjati/smooth-infinite-zoom.