Zero nans — Noobai-based SDXL vPred Merge
A lineart-focused, sketch-style SDXL model merge built on NoobaiXLNAIXL_vPred1, enhanced with handpicked models like AozoraXL, LHC vPred, Duchaiten Style 18 + others, and a self-trained LoRA.
(Edit: use modelsamplingdiscrete if you're using ComfyUI)
🔧 Recommended Settings
Sampler:
Eulercfg++Steps:
28CFG:
1
🖊️ Prompting Tips
Use Danbooru tags & artist tags, check out their wiki on https://danbooru.donmai.us/wiki_pages/help:home
🧬 Merge Components
NoobaiXL, AozoraXL, LHC vPred, Duchaiten Style 18, DimSum, Self-trained LoRA
Description
Not better than v1.5, but slightly different
FAQ
Comments (5)
Doesn't work at all on Reforge. This is the only checkpoint (all current versions) I've encountered that doesn't work on Reforge. Could be my install, but I've just never had this problem with any other checkpoints, and I've used probably over 100 of them at this point.
Not sure if you're interested in a log, I do have one I can get.



