【单人角色出图全流程工作流 / Anima & SDXL 双模型一键切换】
International users, please use the v1.3 English UI.
v1.3
修复了我个人在使用及测试时所遇到的各种问题，大概就是最终版了，请直接下载该版本
v1.2
修复了WD反推不起效的问题以及在参数页面增加了二次采样的步数与CFG值
v1.1
修复了提示词切换在一定情况下会导致D站画廊的附属节点不起效与一些遗留问题
一套面向单人角色立绘的全流程 ComfyUI 工作流，能够实现在Anima模型与传统SDXL系（Illustrious/Noob）模型之间一键切换(或同时使用)的同时保留模型参数。一采与二采(二采通过开关进行控制)可各自独立选用不同底模与 LoRA。
进阶流程：anima文生图（一采）→ ControlNet 深度/姿态控制（可选）→ illu图生图精炼（二采 可换模型 可禁用）→ Ultimate SD Upscale 分块放大(可选是否使用tile模型约束提示词) → 面部/手部 FaceDetailer 自动精修
考虑到生图的多种情况 该工作流内置了通过简单的开关控制送往SD放大的图像 如一采进SD 二采进SD 通过加载图像进SD 通过加载图像进面部/手部细化 均只需要简单的几个开关控制
注:个人comfyui运行及测试的版本为v0.24.1
该工作流特色：
常用参数与模型选择均收纳进可视化参数面板，调参不用进节点图找连线
支持 WD14 反推、Danbooru 取图、LLM API 润色提示词，半自动生成提示词
内置详细中文说明，打开工作流即可查看使用指南哦Ciallo～(∠・ω< )⌒★
显存清理节点保障长流程稳定运行
工作流较复杂，依赖较多自定义节点，首次使用请用 ComfyUI Manager 的"Install Missing Custom Nodes"一键安装或通过下方的插件列表或123云盘下载~
123云盘链接:https://1826369123.share.123865.com/123pan/2dpRjv-B5vX?pwd=0721#
【Single-Character Image Generation Full Workflow / Anima & SDXL Dual-Model One-Click Switch】
v1.3
Fixed the various issues I personally ran into during use and testing — this is essentially the final version, so please go ahead and download this version directly
v1.2
Fixed the issue where WD14 tagging was not working, and added step count and CFG value controls for the 2nd pass to the parameter panel
v1.1
Fixed an issue where switching prompts under certain circumstances could cause the subsidiary nodes of the Danbooru Gallery to not take effect, along with some other legacy issues
A full-pipeline ComfyUI workflow for single-character illustration. Switch between Anima and traditional SDXL-based models (Illustrious/Noob) with a single toggle — or even use both at once — while keeping each model's own parameters intact. The 1st pass and 2nd pass (toggle-controlled) can each independently use a different checkpoint and LoRA.
Advanced pipeline: Anima text-to-image (1st pass) → ControlNet Depth/Pose Control (optional) → Illustrious image-to-image refinement (2nd pass, model-swappable, can be disabled) → Ultimate SD Upscale tiled upscaling (optionally constrain with a Tile-model prompt) → Face/Hand FaceDetailer auto-refinement.
To cover all the different generation scenarios, the workflow lets you route images into SD Upscale or Face/Hand detailing with a few simple toggles — whether it's the 1st-pass output, the 2nd-pass output, or an externally loaded image going into Upscale, or an externally loaded image going straight into Face/Hand detailing.
Note: personally tested and run on ComfyUI v0.24.1.
Workflow highlights:
Common parameters and model selections are all collected into visual control panels — no need to dig through node connections to adjust settings.
Supports WD14 tagging, Danbooru gallery import, and LLM API prompt refinement for semi-automatic prompt generation.
Includes detailed in-graph notes in English — just open the workflow to see the usage guide, Ciallo～(∠・ω< )⌒★
VRAM cleanup nodes keep long pipelines stable.
This workflow is fairly complex and relies on a number of custom nodes. For first-time use, please use ComfyUI Manager's "Install Missing Custom Nodes," or check the plugin list below, or download via the 123 Cloud Drive link (mainland China users) below.
123 Cloud Drive link: https://1826369123.share.123865.com/123pan/2dpRjv-B5vX?pwd=0721#
(International users: the link above requires a mainland China phone number to access — please just use the GitHub links below instead.)
———————————————————————————————————————————
插件列表 Plugin list:
https://github.com/rgthree/rgthree-comfy
https://github.com/ltdrdata/ComfyUI-Impact-Pack
https://github.com/ltdrdata/ComfyUI-Impact-Subpack
https://github.com/ssitu/ComfyUI_UltimateSDUpscale
https://github.com/pythongosssss/ComfyUI-Custom-Scripts
https://github.com/pythongosssss/ComfyUI-WD14-Tagger
https://github.com/yolain/ComfyUI-Easy-Use
https://github.com/Fannovel16/comfyui_controlnet_aux
https://github.com/kohya-ss/ComfyUI-Anima-LLLite
https://github.com/zml-w/ComfyUI-ZML-Image
https://github.com/weilin9999/WeiLin-Comfyui-Tools
https://github.com/Aaalice233/ComfyUI-Danbooru-Gallery
https://github.com/alexopus/ComfyUI-Image-Saver
https://github.com/heshengtao/comfyui_LLM_party
使用模型如下(cn默认使用姿态与深度控制)
Models used below (ControlNet defaults to Pose + Depth control)
面部/手部与反推模型可根据自身需求选择在设置好的控制面板处更换
Face/hand detection and tagger models can be swapped freely in the control panel based on your needs.
光辉姿态与深度控制 Illustrious Pose & Depth Control
Illustrious-XL ControlNet Openpose – v1.0: https://civarchive.com/models/1359846/illustrious-xl-controlnet-openpose
Illustrious-XL Depth Midas – v2.0: https://civarchive.com/models/1023507/illustrious-xl-depth-midas
SD 放大常出现无中生有等问题，tile 能极大缓解这个现象
SD Upscale often results in issues such as creating artifacts out of thin air; Tile can significantly mitigate this problem.
光辉(illu) tile模型 Illustrious Tile model
Illustrious-XL Tile – v1.0: https://civarchive.com/models/1102581/illustrious-xl-tile
NOOB tile模型 NOOB Tile model
Eugeoter/noob-sdxl-controlnet-tile: https://huggingface.co/Eugeoter/noob-sdxl-controlnet-tile