1. Workflow Purpose

This pipeline specializes in video outpainting using Alibaba's Wan2.1-VACE technology, enabling:

• Aspect ratio conversion (e.g. 16:9→21:9)

• Dynamic content generation (explosions/effects)

• Multi-resolution support (480p to 4K)

2. Technical Highlights

• Key Nodes:

  • WanVideoVACEEncode: Video feature encoder with mask support

  • ImageResizeKJ: Lanczos resizing with 16px alignment

  • WanVideoExperimentalArgs: Controls UNet block swaps

• Performance:

  • BF16 precision for VRAM optimization

  • 29-frame batch processing

3. Node Connections

Core Data Flow:
LoadVideo → MaskGeneration → VACE Encoding → Sampling → VideoCombine

4. Customization Guide

To modify output:

1. Change prompt in String Literal node

2. Adjust padding in ImagePadKJ (default 128px)

3. Set target resolution in GetImageSize+

5. Hardware Recommendations

• Minimum: RTX 3090 (24GB VRAM)

• Optimal: RTX 4090 + --xformers

• Storage: SSD recommended for video caching

6. Expected Output

Sample Result:

• 16fps MP4 video with expanded content

• Preserved original motion dynamics

• Generated elements blend seamlessly