1. Workflow Overview

This workflow generates high-quality character images and animates them with 360° rotation. Key features:

• Character Generation: Creates realistic/anime characters from input images or text prompts.

• Video Synthesis: Converts image sequences into smooth rotation animations.

• Post-Processing: Includes upscaling, frame interpolation, and face detailing.

Core Models:

2. Key Nodes & Installation

Critical Nodes:

1. WanVideoModelLoader

   • Purpose: Loads base video models (e.g., Wan2.1-I2V-14B).

   • Install: Requires ComfyUI-WanVideoWrapper plugin via ComfyUI Manager.

   • Dependency: Download models from NetEase and place in models/wan_video.

2. WanVideoLoraSelect

   • Purpose: Applies rotation LoRA (e.g., 360_rotation_lora_flux).

   • Install: Place LoRA files in models/loras.

3. GIMMVFI_interpolate

   • Purpose: Frame interpolation for smoother videos.

   • Install: Clone ComfyUI-GIMM-VFI from GitHub.

4. FaceDetailer

   • Purpose: Auto-refines facial details.

   • Install: Built-in node; requires UltralyticsDetectorProvider (YOLO model).

3. Workflow Structure

Group 1: Character Generation

• Inputs:

  • Image (via LoadImage) or text prompt (e.g., "1girl, fashion model").

  • Parameters: Resolution (default 512x768), seed, steps (20).

• Output: Generated character image (PNG).

Group 2: 360° Rotation Animation

• Inputs:

  • Generated image + rotation LoRA.

  • Parameters: FPS (16), rotation control (via WanVideoBlockSwap).

• Output: Rotation video (MP4).

Group 3: Post-Processing

• Nodes:

  • CR Upscale Image: 8x upscaling.

  • GIMMVFI_interpolate: Frame interpolation to 32fps.

  • VHS_VideoCombine: Final video rendering.

4. Inputs & Outputs

Input Parameters: （json）

{
  "prompt": "1girl, Asian fashion model",  
  "negative_prompt": "low quality, blurry",  
  "seed": -1,  // Random seed  
  "width": 512,  
  "height": 768  
}

Output:

• Video file: /ComfyUI/output/XIAO_xxxx.mp4 (H.264).

5. Notes

1. Model Dependencies:

   • Download Wan2.1 models (~14GB); VRAM ≥16GB (24GB recommended).

   • Missing LoRA triggers LORA not found; verify filenames.

2. Optimization:

   • Run ComfyUI with --highvram flag.

   • Disable unused Groups (e.g., character generation) to save resources.

3. Troubleshooting:

   • Resolution mismatch: Use ImageResizeKJ to unify dimensions.

   • VRAM overflow: Enable PurgeVRAM to clear cache.