<h3 id="note">NOTE</h3>This is eps model (epsilon-prediction) which most model use. I don't know how to make vpred (V-prediction) that most on NoobAI model and most of time it just gave me headache so I just stick with eps and I prefer this over vpred.<h3 id="about">ABOUT</h3>NoobAi Beast Mix XL is like an upgrade model for <a target="_blank" rel="ugc" href="https://civarchive.com/models/496600/pony-beast-mix-xl">PonyXL model</a> version for <a target="_blank" rel="ugc" href="https://civarchive.com/models/496600/pony-beast-mix-xl">AI Beast Mix</a> . Similarly, this is a model merge to achieve semi-realism/ Realism result for furry character with NSFW kink such as muscle or hyper.I personally really happy with NoobAI with how much better it works. Its like an upgrade from Pony to NoobAi. The prompt also works better and generation pretty much close to Pony without the "score" prompt.As usual I recommend using the model with Hires.Fix and Inpaint on certain part to fix the visual.<h3 id="lora">LORA</h3>I notice this model support both Illustrious Lora and Pony lora. They're SDXL model to begin with but NoobAI handle those lora lot better than them handling other loras.<h3 id="recommend-base-prompt:">Recommend Base Prompt:</h3>Positive : masterpiece, best quality, good quality, realistic, furry,Negative : normal quality, worst quality, old, early, signature, watermark, comic, cartoon, human<h3></h3><h3 id="generate-setting">Generate Setting</h3>Sampling Method: Euler ASchedule type: KarrasSampling steps: 24Width X Height: 576 x 768CFG Scale: 4 / 2<ul><li>4 is the default CFG scale for this model. If go higher there's chance the image become too saturate. Sometime even at 4 can be saturate, so you need to reduce it slowly to 3 and then 2. Try experiment it.</li></ul>Hires. fix (reccomended)Upscaler: LatentHires Steps: 20Denoising Strength: 0.5-0.6Upscale by: 1.34

NoobAi Beast Mix XL

Initial Release, vae already baked in

Files with SHA256: ab2067c28886fc70255b42a55109a8c71d6175a4b67f062cc4b4f055d5155f12