Experimenting with an end-to-end AI character workflow in #ComfyUI! 💡 Character Sheet -> 3D Model.

(A-Z why For production system (api))

Here's the breakdown:

1️⃣ Image-to-Prompt: Using the brilliant ComfyUI-OllamaGemini node by al-swaiti (https://github.com/al-swaiti/ComfyUI-OllamaGemini) to analyze an input character image and automatically generate a detailed text prompt.

2️⃣ Character Sheet Generation: Feeding that generated prompt into an image generation model to create a character reference sheet. The key here is achieving a usable sheet layout without relying on ControlNet or explicit pose workflows for this stage!

3️⃣ 2D-to-3D Conversion: Taking the generated 2D character sheet image and converting it into a 3D asset using the open-source #Hunyuan3D 2.0 model.

Want to try the 3D part?

• Download the model.fp16.safetensors from tencent/Hunyuan3D-2mv (specifically the hunyuan3d-dit-v2-mv-turbo version) on Hugging Face.

• Rename it to hunyuan3d-dit-v2-mv-turbo.safetensors.

• Place it in your ComfyUI/models/checkpoints directory.

This pipeline showcases the power of chaining different AI models in a modular environment like ComfyUI for complex creative tasks. From visual analysis to 3D generation!