<h2 id="about-this-workflow">About this Workflow</h2>This workflow allows you to generate detailed Character Design Sheets from a single reference image. These sheets can later be used as reference frameworks for video generation models like Seedance 2.5 or MiniMax Hailuo H3.In my experience, Qwen Image 2.1 is the first open-source image editing model capable of natively generating clean, highly detailed character design sheets right out of the box.<h2 id="recommended-settings">Recommended Settings</h2><ul><li>CLIP Encoder: Use FP32 / FP16 / BF16 precision for maximum image quality and prompt adherence. Tests show that INT8 and other lower-bit quantizations introduce noticeable artifacts and reduce image clarity. Don't skimp on SSD space — install <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_bf16.safetensors">qwen3vl_8b_bf16.safetensors</a>.</li><li>Sampler &amp; Scheduler: Use res_2m + beta to reduce artifacts and improve detail. While I haven't run extensive benchmarks on every combination, this pairing yielded the best results.</li><li>Resolution (Megapixels):<ul><li>3.4 MP — Sweet spot for speed and clarity.</li><li>6.0 MP — Superior quality and detail, though generation time increases significantly.</li></ul></li></ul><h2 id="how-to-use">How to Use</h2><ol><li>Install custom nodes via ComfyUI Manager:<ul><li><a target="_blank" rel="ugc" href="https://github.com/kijai/ComfyUI-KJNodes">ComfyUI-KJNodes</a></li><li><a target="_blank" rel="ugc" href="https://github.com/lunaaispace-eng/ComfyUI-DeGrid">ComfyUI-DeGrid</a></li><li><a target="_blank" rel="ugc" href="https://github.com/goodguy1963/ComfyUI-ThinkingLLM">ComfyUI-ThinkingLLM</a> (you can replace this with the standard Generate Text node if you want to use official Qwen Image 2.1 PE weights). Don't forget to install llama-cpp-python as specified <a rel="ugc" href="https://github.com/goodguy1963/ComfyUI-ThinkingLLM/blob/main/docs/LLAMA_CPP_PYTHON_VISION_INSTALL.md">here</a>.</li></ul></li><li>Download GGUF PE I2I weights and place them into <code>ComfyUI/models/LLM/GGUF/</code>.<blockquote>Important: Both the main PE model and the mmproj file must be present in this folder, otherwise <code>ComfyUI-ThinkingLLM</code> will not detect it.</blockquote></li><li>Run generation: Load your reference image, select the path to the PE model, and start generating.</li></ol><h2 id="useful-resources">Useful Resources</h2><ul><li>Official Qwen Image 2.1 Weight for ComfyUI: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1">HuggingFace</a></li><li>GGUF PE T2I: <a target="_blank" rel="ugc" href="https://huggingface.co/prithivMLmods/Qwen-Image-2.1-PE-T2I-GGUF">HuggingFace</a> | GGUF PE I2I: <a target="_blank" rel="ugc" href="https://huggingface.co/prithivMLmods/Qwen-Image-2.1-PE-I2I-GGUF">HuggingFace</a></li></ul>

Files with SHA256: 86224ce45ab9e3aaa27ac2f45bfffd337287d1be96c10b97780f1bbd3c67ef48