The ultimate goal of this model is to provide an experience very similar to the already fairly competent base of NoobAI v-pred, while fixing up rough edges. Many other merges suffer from the bimodality of either having good prompt adherence (closer to base noob) or good default aesthetics (closer to illustrious).
Ideally, both can be encapsulated in a model without sacrificing too much model knowledge to achieve this.
Recommended Settings
For optimal results, we recommend the following inference parameters:
Sampler:
EulerorEuler AScheduler:
NormalorBetaSteps:
16-24CFG Scale:
3-6Resolution:
For general use:
832x1200(or similar aspect ratios with a total area around 1024x1024)For V9.1 and above: Can natively handle
1536xresolutions.
Description
Small tweak to the model to make it more stable; CFG 3 recommended. LoRA compatibility should be better, styles respond slightly better, default aesthetic may be a little hit or miss.
FAQ
Comments (5)
For a bit more clarity regarding the 'hotfix' version; I think I screwed up a bit of the workflow.
The text encoder should now primarily be noob vpred, and some model layers were merged from base noob to fix some quality and anatomy issues. Overall, the model is more stable, sharper, and generally better than the regular 9.2 version, apologies 🫡
Also LoRAs trained on Noob vpred and 1.1 (eps) should work a bit better than 9.2.
I still personally think this is a slight downgrade in quality (default aesthetic, anyway) from 9.1, but a step in the right direction overall.
Basically this is NoobAI re-trained for better furry knowledge.
This is honestly a damn good choice for a base model when merging (though I havent tried, but I might soon)
It gens images great already. But if people started used Illustrious 2.0 + Wahtastic for furry merge basesI think it would totally change the landscape for furry ILX models
Also, still a better trained model than pony v7
It's a great model. Comparing it to my other regulars this one seems to do a little bit better in a lot of furry/feral specific areas. It does native highres pretty well. The only real thing that's off is the colors/darkness is a bit brighter compared to the others I use. And LORAs might be a bit touchy.
V10 soon! I think I've got it in a nice spot :)
Images linked below:
Taken from the latest experimental on hf:
https://files.catbox.moe/c6f358.png
https://files.catbox.moe/v0u3j4.png
I'll be waiting patiently
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.