<aside> 💡 Note that on every model except the Default one, the first loading time will take 1 to 2 minutes. You only need to wait once, after which image generation is quick (seconds).

</aside>

Default (Stable diffusion XL)

This model is good at generating pretty much anything but doesn’t excel at anything specific. The big plus is that you can use any of the pre-built styles:

It is also the only model that currently supports image to image.

Make sure to activate the “rephrase with GPT” option of the sidebar to get great results with basic prompts.

Undraw icons

This model is a sample to show you how you can train models to replicate styles. It was trained with 50 images from https://undraw.co/illustrations, well captioned.

Ask it to generate anything with a prompt such as:

Photorealistic (epic realism)

This model is great at generating realistic photos of people, objects and places - but has no style variation and is quite specific.