1. Workflow Overview

This WAN2.1 I2V workflow specializes in transforming static figurine images into animated videos. Key features:

• Input: A PNG image (e.g., ComfyUI_temp_sydys_00036_.png) → Output: 3D rotation/transformation video

• Supports background preservation and motion control (via SLGArgs node)

• Exports MP4 video (720P, 16FPS by default)

2. Core Models

3. Key Nodes

4. Workflow Groups

1. Model Loading Group

   • Nodes: LoadWanVideoT5TextEncoder, WanVideoVAELoader

   • Input: Model paths (e.g., umt5-xxl-enc-bf16.safetensors)

   • Output: Initialized encoders + VAE

2. Image Processing Group

   • Node: WanVideoImageClipEncode

   • Input: 1024x1440 PNG image (transparent background recommended)

   • Output: Image embeddings

3. Video Generation Group

   • Nodes: WanVideoSampler + WanVideoDecode

   • Params: Seed 34660692369907, Steps 20

   • Output: Latent video sequence

4. Video Export Group

   • Node: VHS_VideoCombine

   • Params: FPS 16, CRF 19

   • Output: WanVideo2_1_00001.mp4

5. Inputs & Outputs

• Inputs:

  • Image: 1024x1440 PNG (transparent background)

  • Prompt: "A girl rotates 360° and transforms into a 3D figurine"

  • Negative Prompt: "Low quality, background change, distorted limbs"

• Output:

  • 720P MP4 video (H.264, YUV420P)

6. Notes

1. Hardware Requirements:

   • VRAM ≥16GB (due to bf16 + fp8_e4m3fn optimization)

   • Launch with --highvram flag

2. Troubleshooting:

   • Model Path Error: Ensure .safetensors are in ComfyUI/models/wanvideo/

   • Video Flickering: Adjust WanVideoTeaCache parameter 0.3