Task:

Develop a ComfyUI workflow that allows users to easily utilize the Kandinsky 3.1 (not older versions like 2.2) model for image generation.

Requirements:

• The workflow should be well-structured, intuitive, and clearly documented.

• It should include nodes for essential functions such as:

  • Loading the Kandinsky 3.1 model

  • Setting image dimensions

  • Inputting text prompts

  • Adjusting generation parameters (sampling methods, steps, CFG scale, etc.)

  • Outputting generated images

• The workflow should be compatible with the latest version of ComfyUI.

• Bonus points for incorporating advanced features like:

  • Prompt beautification feature

  • KandiSuperRes

  • Inpainting/outpainting functionality

  • Kandinsky IP-Adapter & Kandinsky ControlNet

Supplemental Materials:

• Kandinsky 3.1 Source Code: https://github.com/ai-forever/Kandinsky-3

• Hugging Face Model: https://huggingface.co/ai-forever/Kandinsky3.1

• Kandinsky's "AUTOMATIC1111" (Kubin): https://github.com/seruva19/kubin

• FusionBrain Testing Website: https://www.sberbank.com/promo/kandinsky/

• Outdated plugin for 2.2: https://github.com/vsevolod-oparin/comfyui-kandinsky22