Welcome to OS Models!
This is my second true checkpoint merge, and the goal is simple — pure 2D cel-shaded anime with MUSCLES. This checkpoint is built from the ground up with one vision in mind: that clean, crisp, authentic anime screencap look you'd pull straight from your favorite animated series. Sharp linework, bold flat cel shading, punchy colors, and that unmistakable hand-drawn 2D feel that makes anime so iconic.
It handles a wide range of anime aesthetics too — bold shonen energy, clean slice-of-life warmth, sharp seinen contrast, and more — all while keeping that flat, 2D cel-shaded core. LoRA compatibility has been strong in testing, so your favorite style and character packs should slot right in.
This is my first release after a lot of testing and refinement, so please give it a go! Drop your results, questions, and suggestions in the comments — I'd love to see what you create.
Donations to Yellow Buzz are hugely appreciated and go straight back into building more LoRAs and checkpoints for everyone to enjoy!
Positive: high_quality, best_quality, highres, masterpiece, anime screencap, 2D, cel shading, flat shading, clean lineart, bold outlines,
Negative: (lowres:1.2), (worst quality:1.4), (low quality:1.4), (bad anatomy:1.4), realistic, photorealistic, 3D render, soft shading, painterly, blurry lineart, multiple views, jpeg artifacts, artist name, censored, (abstract:0.91)
Sampler & Scheduler Suggestions:
Euler Ancestral + Normal/SGM Uniform — 30 steps @ 4.5 CFG
DPM++2M + Karras — 30 steps @ 4.5 CFG
This checkpoint includes a built-in VAE and supports hi-res fix.
Hi-Res Suggestion: Euler or DPM++2M — 12–15 steps @ 0.35 denoise (1.5× or 2×)
Description
Merge 1: Yiffy Mix 6.4 (0.6) + ReedFurryMix SDXL 2.4 (0.4)+ Anime screencap lora (0.6) + a few muscle loras and one detail lora = Merge 1
Merge 2: Merge 1+ Nova Anime XL 1.9 (0.3) = Final Merge
Comments (1)
First of all, thank you so much for your work on this model. I greatly appreciate the intention behind it. Personally, my Graal is to find a checkpoint that consistently generates images with flat colors, without any gradients. So, dark colors for shadow areas and light colors for highlights but not gradient at all between them for volume. I have yet to find such a checkpoint tho.







