v2: Same thing but with the new 0.75s version.
v1: Merge of NoobAI-XL-Vpred-v0.65-Cyberfix and EasyFluffXLVpredIter. Compared to the base model, there are less artifacts and backgrounds have more detail.
This model uses V-prediction; enable advanced sampling in reforge (dev branch), or use the ModelSamplingDiscrete node in comfyui.
Usage tips + (official user guide)
samplers: euler / dpmpp_sde (~30 steps)
cfg: ~4 + enable rescalecfg at 0.7
positive prompt: very awa, artist names, character/copyright tags, general tags
negative prompt: text, 2koma, spot color, scan, greyscale, abstract, anthro, yellow theme