Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
我感觉在zimage部分甚至第一步klein就直接开始进行去衣效果并不好,可以之后再用klein单独进行去衣。 I feel that starting the undressing process directly in the zimage section, or even at the first step with klein, doesn't yield great results. We can use klein...
Runtime profile
Source description
我感觉在zimage部分甚至第一步klein就直接开始进行去衣效果并不好,可以之后再用klein单独进行去衣。 I feel that starting the undressing process directly in the zimage section, or even at the first step with klein, doesn't yield great results. We can use klein separately for that step later.
1.5更新 进行了默认数据和提示词的优化,可以根据需要自行修改。 I've optimized the default data and prompts. Feel free to tweak them to fit your needs.
1.4更新 分别给了zimage和klein一个joytag节点,为了区分筛选强度。 I assigned separate JoyTag nodes to zimage and klein to differentiate their filtering strengths.
1.3更新 添加文本连接节点,joytag连接zimage,可根据需要连接klein。 Add a text concatenation node. Connect JoyTag to ZImage, and optionally connect Klein if needed
1.2更新 添加joytag节点
1.1更新 使用图层缩放那个节点代替整数运算 Used Layer Scale instead of Integer Math
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Use Klein+Zimage二转三anime to real to turn an input anime image into an anime-to-realism image through ZImage and Klein; provide text prompts or JoyTag text, and produce saved JPG images or a stitched long scroll/strip image.
The workflow combines ZImage and Klein stages, with JoyTag text feeding ZImage and optionally Klein.
It includes separate JoyTag nodes for ZImage and Klein, plus image stitching for long scroll or strip compositions and JPG saving.
Provide an image and prompt text; the workflow can concatenate text, then save JPG images or combine images into a long scroll or strip.
It is based on Flux.2 Klein 9B and published by Blackbea; the SetInt and VQA nodes remain disabled but can be reconnected if needed.
Custom node packs listed are comfyui_fearnworksnodes, Comfyui_joytag, comfyui_memory_cleanup, ComfyUI-QualityOfLifeSuit_Omar92, and comfyui-various.
Model files listed include Flux2 Klein动漫转写实真人 AnythingtoRealCharacters.safetensors, flux2-vae.safetensors, flux2Klein9BTrue_v10Fp8.safetensors, qwen_3_4b.safetensors, and qwen_3_8b_fp8mixed.safetensors.
The list also includes Qwen-Image-Lightning-8steps-V1.0.safetensors, Z-image-vae.safetensors, and ZZZ黑兽darkbeastz6BlitzScaled.CEhH.safetensors.
Edit the default data and prompts to fit your needs.
Suggestion · not verified
Use the Layer Scale control to adjust image dimensions; the description notes that it replaced Integer Math.
Suggestion · not verified
For a later upscale, the described sequence is SEEDVR2 to 3k, followed by SDXL for final high-definition upscaling.
Suggestion · not verified
Before use, verify that Z-image-vae.safetensors, Flux2 Klein动漫转写实真人 AnythingtoRealCharacters.safetensors, and ZZZ黑兽darkbeastz6BlitzScaled.CEhH.safetensors are available under the exact listed names.
Suggestion · not verified
Before loading flux2-vae.safetensors or qwen_3_4b.safetensors, verify the exact files you plan to use.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
1 excerptSource context: 641 downloads · Type Workflows · Base model Flux.2 Klein 9B
贝露镇楼。Vell Sain to bless this thread
修改自千问+ziamge的二次元转真人工作流,将千问部分替换为站内的klein二转三工作流 https://civitai.com/models/2345939/flux2-klein-the-ultimate-anime-to-realism-workflow-8gb-vram-friendly ,并且添加了整数运算节点调整图像尺寸,添加清理缓存,添加权重lora加载器,添加jpg保存节点,添加卷轴图合并。无效了原先的setint节点和VQA节点,可以重连。
后续高清放大,推荐使用SEEDVR2直接放大到3k,然后用SDXL进行放大。
Anime-to-Realism Workflow Modification
Base Workflow: Modified from the "Qwen + zimage" anime-to-realism workflow.
Key Changes & Features:
Core Replacement: Replaced the Qwen (VLM) section with the Klein anime-to-realism workflow found on Civitai: Flux2 Klein - The Ultimate Anime to Realism Workflow .
Resolution Control: Added Integer Math nodes to adjust image dimensions.
Optimization: Added Cache/VRAM Clearing nodes.
LoRA Support: Added a LoRA Loader (with weight adjustments).
Output: Added a Save as JPG node.
Composition: Added an Image Stitching/Merging node (for creating long scroll/strip images).
Notes on Nodes:
The original SetInt and VQA nodes have been disabled (muted) but are left in the workflow; you can reconnect them if necessary.
Upscaling Recommendation:
Step 1: Use SEEDVR2 to upscale directly to 3k resolution.
Step 2: Follow up with SDXL for the final high-definition upscale.
Estimated VRAM requirement
Estimate unavailable
16.5 GB across 2 of 8 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
13 requirementsFlux2 Klein动漫转写实真人 AnythingtoRealCharacters.safetensors
Not resolvedLoRA · Unknown
VAE · Hugging Face · Comfy-Org/flux2-dev
Checkpoint · 8.46 GB · SAFETENSORS · fp8 · pruned · Unknown
Text encoder · 8.07 GB · SAFETENSORS · fp16 · Hugging Face · Comfy-Org/flux2-klein-9B
Qwen-Image-Lightning-8steps-V1.0.safetensors
File unverifiedLoRA · Hugging Face · lightx2v/Qwen-Image-Lightning
Z-image-vae.safetensors
Not resolvedVAE · Unknown
ZZZ黑兽darkbeastz6BlitzScaled.CEhH.safetensors
Not resolvedCheckpoint · Unknown
Node pack · Registry
Node pack · Registry
Node pack · Registry
Node pack · Registry
Node pack · Registry
