文件清单 线路:原始(huggingface.co) · 切换
下载后放进对应的 ComfyUI 目录。models/ 指 ComfyUI 的模型根目录。
说明
RDBT | MOSS Cauldron
作者:reakaakasky
RDBT | MOSS Cauldron
(TBD) This page contains merged ckpt, in ComfyUI int8 format (converted by comfy-kitchen). Uploaded here mainly for convenience. So you won't have to stack those LoRAs directly on int8 ckpt (which means de-quantizing and re-quantizing twice). For advanced users: If you can make the int8 ckpt on your own device, you don't need to waste time and download the ckpt here. Recipes are below. About my RDBT dataset, as well as original LoRA files, see the main page for more info. Versions/recipes: v0.4 (Kromoss + RDBT + DMD LoRA) (test):
- Base model: kromoss_bf16_OPD-priming.safetensors
- RDBT v0.4 that is trained against the OPD model. Then strengthen the default anime style with a "official art" dataset. This dataset contains many samples from "azur lane". Body shape might be more curvy. Because this is a test lora against a test base model, I won't release the LoRA file publicly.
- DMD: opd-priming-8-steps-dmd-lora_cfg-static_20000.safetensors
- Trigger words: "Anime-style illustration of ..."
- Note: use 8~12 steps, CFG 1, euler sampler. Check kromoss model page for more info.
NOTICE: RDBT is by reakaakasky. Sharing merges using RDBT model/lora is not allowed. This "restriction" won't affect anyone who use it. It's only aimed at those who steal others' models to sell. Special trigger words and latent watermark included. Kromoss is by lodestones, released under the Apache License, Version 2.0. Built on MOSS-Image by the OpenMOSS Team.
