This is a Diffusers version of Rev Animated version 1.2.2 specifically for InvokeAI users. The original checkpoint with all the info on the model can be found here: https://civitai.com/models/7371/rev-animated

The differences include:

  • Changed the scheduler to EulerAncestralDiscreteScheduler, works very well with Euler_A.
  • Updated the text_encoder config.json file setting num_hidden_layers to 11 instead of the default 12. This is the equivalent of 'Clip Skp 2' in Auto1111 as I understand it and have tested it. Please let me know on the InvokeAI Discord if you encounter issues.
  • Don't expect it to build the exact same image from the exact same seed as you would in Auto. Invoke and Auto are very different from another, and diffusers are even more different. You should, however, get some awesome images.
  • I'm sharing this because diffusers are amazing and I think more people should use them. ;)
  • Comes with the default vae used during the conversion into diffusers format. The original author recommends a few different vaes at the link above, I wanted you to be able to mix and match. The examples below were made without a vae.

Here are a few example images:

Example Image Example Image Example Image

Downloads last month
305

Spaces using LottePeisch/RevAnimated-Diffusers 50