Xavier Void Fused Anima – v1.0
(Read the Version Tab for full Detail about the models)
I really love XAVIER and Indigo Void Furry Fused XL, so I created a merge that combines them both , Since Xavier dont have a Anima model(as v2 coming out) i merged the models that made Base Xavier and combined it with v1 to get xavier void fused anima
Anima Merge v2:
Xavier Void V1
Homosimile
Indigo Furry
Anima Merge v1:
Mol_keun Anima base
Added custom xavoid style LoRA
Goal & Aesthetic:
My goal was to blend Xavier’s strong fantasy style and character knowledge with Indigo Void’s semi-realistic 2D/2.5D aesthetic and clean detail work. The result is a flexible model in both semi-realistic and fantasy outputs capable of both SFW and NSFW (a bit more oriented toward NSFW).
Images in Checkpoint Cover Image Settings (mostly) :
Sampler - Euler a
Steps - 28 Steps
CFG Scale - CFG 5
Recommended Settings :
Sampler - Euler a / DPM++ 2M Karras (User Preference)
Steps : 25 - 40
CFG : 4 - 8
High-Res Fix (if needed✔️)
Positive tags
masterpiece, best quality, score_7,Negative tags
worst quality, low quality, score_1, score_2, score_3, blurry, jpeg artifacts, sepia, text, logo, shiny skin, shiny clothing, signature, username, bad hands, mutated handsExperiment different settings and tags to your Generating style
Credits
Huge thanks to the amazing model creators:
@ok_furry_image
@r12x5s
@Mol_Keun
@Kairen92
Please check out their original work as well.
Made with love ❤️
Description
this is a much more improved version of xavoid , since theres no xavier anima i made a merge that mimic xavier as much as possible by adding homosimile and indigo furry mix (both anima) to base xavier void anima (which is molkun plus custom xavier style)
FAQ
Comments (3)
is there a modified text encoder yet for Anima Furry models that also has better understanding? I feel like with a lot of models I have a hard time getting things like "canine P****" and im wondering if its not cuz of the text encoder instead of just model training data?
I think that particular anatomy was just poorly trained from the start since things like horse anatomy are well preserved. Training the text encoder really isn't recommended for anima because it can rapidly degrade other stuff. I've successfully trained loras to better represent that particular anatomy and it works fine (I'm not good at separating style from concept though so it's not super useful for other people), we just need people who know how to train well to actually train it and get it into the wild so it gets absorbed into these merges.
@sansmia I forget sometimes that most people merge instead of train cuz it can be "eaiser" at times (myself included xD). Well, here's hoping for a very good furry train someday, and hopefully without using LoRA's in the training. I heard that can sometimes ruin stuff on the model internally like it did apparently with some SDXL models.