ModelExMix XL - General Anime Base Merge is a three-way checkpoint merge in the Illustrious ecosystem.
A balanced, general-purpose anime checkpoint built for everyday use: stable anatomy, clean lineart and restrained coloring that does not fight your prompt. School, slice-of-life, urban and fantasy scenes all land without extra tuning, and it takes LoRAs cleanly.
Recipe: 0.45 x Model-EX v7.0 (Illustrious) + 0.30 x WAI-illustrious-SDXL v17 + 0.25 x Illustrious-XL-v1.0, applied to the UNet only (VAE and text encoder inherited from the first model).
What you get:
Stable anatomy and clean, readable lineart
- Coloring tuned for 综合 results without extra tuning
- Reliable prompt following and clean LoRA stacking
[Recommended settings] Sampler DPM++ 2M or Euler a, 28-32 steps, CFG 7, 832x1216. Danbooru-style tags work best. Use bf16 where possible - fp16 can soften output.
[License] CreativeML Open RAIL-M, inherited from the base models. Please credit the base models if you redistribute. Part of the merged-checkpoint series.