A prototype Lora trained on 100+ High-quality "Style Hacked" NovelAI (NAI Furry-3) Generations to enable high-quality baseline generations with NoobAI-based models with a smooth-lined 3D "Anime" Style.

This Lora is currently a prototype and has baked-in bias I'm presently generating a much larger dataset to reduce the bias overhead and increase the overall quality and style. For both Anime and Furry Generations.

Trained on NoobAI V1.0 and tested on ChromaXL Mix "Mango" https://civitai.com/models/907792?modelVersionId=1024571

Version 1.5 Update
For the best results, I highly suggest using ChromaXL Mix "Mango" with this Lora. These settings guidelines are intended to use the model "ChromaXL mix Mango" and not NoobAI V1.0 since I haven't had enough time to test NoobAi itself.
Lora Strength 0.1 to 0.7 any higher will start to enforce HEAVY Bias.
CFG 1.7
Euler A with (AYS) Align your steps as the scheduler at 12 steps.
All normal 1K res formats work fine on it. E.g. Square, landscape, portrait.

This is my first attempt at a complex "Style Lora" so expect issues.

Current strengths
-Vastly improves Solo character generation quality on NoobAI Non-Vpred models.
-Works well on nearly all "Furry" anthro species including Existing characters in NoobAi.
-Has higher attention to detail for hands and feet and will generate fewer artifacts.
-Supports a large range of poses without much issue.
-Supports both Male and female generations without many issues.

Current weakness
Depending on Lora's strength, it will bake in Bias towards its training set, for example, Fur colors and/or patterns and Tail types.
-It can struggle on Duo or Trio generations and may generate artifacts or low-quality outputs.
-Its style is hard-baked and does not work well with other "style Lora's" for example "Add detail" and it's not recommended to use this Lora with any other Lora's unless its weight is set very low E.g 0.2 or less
-The Lora struggles to generate "Humans" without weighted tags and the same goes with "Scalie" Anthros you can bypass this issue by increasing the tag weight to 1.2-1.3 or reducing the Lora strength.

This is currently a work in progress and I hope to triple the baseline dataset over the coming days and or weeks to greatly reduce bias and support Duo/Trio content much more smoothly.