The Anima model can effectively render multi-character actions based on natural language prompts, but I’m not satisfied with its visual style. To address this, I’ve added the ill model to make image-to-image. I’ve split the prompts into separate sections for easier and more flexible editing later.

If the generated image appears overly bright after upscaling, there are adjustable parameters to fine-tune the exposure or contrast.

Additionally, I’ve included a “Find & Replace” node to make it easier to swap or modify characters in the scene.

This workflow includes acceleration nodes and Turbo LoRA models for faster performance — but please review the instructions before using it.

Some nodes are not available for direct download from ComfyUI, so I’ve provided corresponding GitHub links. Please visit the GitHub repository to obtain those nodes.

Thank you for using the workflow!