This is really great, @Paul_S. Thanks for linking! Iām also interested in your experience with creating consistent images/characters between runs like we were talking about last night, too. Iāll create a separate thread for that.
I hear this was just āreleasedā (in beta) yesterday or so:
I also heard that it was āethically sourcedā - that the model was trained on āopen sourceā and copyright-expired (public domain - as in true public domain, not just āitās on the internetā) imagery.
Also heard that it uses Stable Diffusion behind the scenes. I have not verified any of these claims, but thought they were worth mentioning.