After extensive testing of the first version, I conclude that the first model will not be further developed. It is too unstable. However, for the few who liked the approach I chose and saw its potential, I am pleased to announce that version 2.0 is almost ready, which I approached much more thoroughly. I see no sense in continuing to specify models as I did before for several reasons - firstly, there are more of them, and secondly, providing full accountability may require a complete list of all models used by other authors??? The list will be endless. Nevertheless, below you can see recommended models, many of which I have used to some extent.
I.S.S.: I've noticed that the number of followers is significantly lower than the number of downloads, which is understandable. However, it's surprising that there are people who appreciate my work but haven't followed me yet. I'm grateful for all those who appreciate my work, but I'd love to see more active engagement from our community. There will be more LoRA, text inversions, and I’ll be regularly updating and improving styles for my models. Plus, you’ll probably be able to peek at the working prompts. So join in the process - don’t forget about the importance of feedback.
Inpaint model works great with 'Poor man's outpainting' script, make sure you use the original prompt, settings and seed.
DDIM sampling is required in general inpaint mode (in case you didn't know).
P.P.S.: It's helps:
catdog: https://drive.google.com/file/d/1cCZgx_znsnnfwXlreHpdnCF1ocSP2wOB/view?usp=sharing
catbird: https://drive.google.com/file/d/1hKfdwcH7BYnNqpwwA2Cnm7lO3biqe1gT/view?usp=sharing