Alright ! So I finally made a Lora for the Revenant. Not perfect by any means, still has a lot of issues, but I'm satisfied enough for a first version.
I'll keep trying to make a better one for sure.
V2 Edit - Now that's a result i'm more satisfied with ! Base style is, I think, less appealing, but that's to be expected when I threw more diverse sources into the dataset. Upside to that is I think a better flexibility on both style adherence and composition !
Description
First version, still has a lot of issues, but good enough for me.
FAQ
Comments (5)
Very nice! I realized the v1 has a more 2d style and the v1.5 version is more realistic, which are both different than your Recluse lora. How did you train them to get differing results like that? Awesome lora though and thank you for doing Revenant! Looking forward to Duchess as well eventually :)
They both use the same settings and dataset. Only difference is the base model. 1.5 is trained on Illustrious 2.0
I did the same on different versions of the Recluse, V4 is Illustrious 2.0 too
oh I see! I've never tried training on different illustrious base models. Illustrious 2.0 is super cool
@Aditai Now, on to the Duchess !
@Aditai Duchess is now available !
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.