If you want to turn simple images into smooth AI animations, Alibaba’s Wan 2.1 InPaint in ComfyUI is a game-changer. I’ve tested this workflow extensively, and in this guide, I’ll show you how to create start/end frame videos—even on low VRAM setups.
Why Wan 2.1 InPaint?
This tool lets you generate videos by analyzing just two images:
- Start Frame (e.g., front view of a car)
- End Frame (e.g., rear view)
The AI fills in the motion between them, perfect for:
✔ Morphing effects
✔ Simple animations
✔ Low-budget projects
Step 1: Install Wan 2.1 in ComfyUI
- Download the Wan 2.1 InPaint model (I’ll link it below).
- Place it in:
ComfyUI/models/diffusion Load it using the “Juan Fun InPaint 2 Video” node.
Step 2: Optimize for Low VRAM
If your GPU struggles:
- Use the 1.3B model (not 14B)
- Set resolution to 512×512 or lower
- Enable –lowvram in ComfyUI startup commands
Tested on a 4GB GTX 1650—works smoothly!
Step 3: Run Your First Interpolation
- Connect your images to CLIP Vision Encode.
- Set:
- CFG 0: Check raw motion (catches errors early)
- CFG 4-5: Final render (better details)
- Generate!
Pro Tip: For natural movement, keep start/end frames similar (same character/lighting).
Fixing Common Issues
❌ Problem: Blurry output
✅ Fix: Upscale with Tile LoRA after generation
❌ Problem: Warped faces
✅ Fix: Add midpoint frames or reduce CFG
❌ Problem: Low VRAM crashes
✅ Fix: Use –medvram or slice video into shorter clips
Free Resources
