Anti-blur / Background Stabilizer LoRA (Wan2.2 I2V)

This LoRA improves background stability and spatial coherence for Wan2.2-based Image-to-Video models.

It focuses on reducing unwanted motion, flickering, jitter, and deformation in the background while keeping the original model’s visual identity untouched.

Extracted directly from the “rapid-mega-aio-v12” UNet-only model by isolating the layers responsible for spatial consistency:

• Self-attention
• Cross-attention
• Feed-forward layers This LoRA does not alter style, color, or character appearance — it purely stabilizes the scene.

Recommended use

• Works with WanVideoSampler / WanVideoModelLoader
• Optimal strength: 0.2 – 0.7 Lower values → subtle stabilization Higher values → stronger anti-blur effect
• Compatible with other LoRAs such as: Lightning LoRA Motion LoRAs Angle Enhancement LoRA Any style/character LoRA

Features

• Reduces background flicker & “breathing”
• Minimizes motion jitter / unwanted shaking
• Improves spatial consistency across frames
• Enhances readability of complex backgrounds
• No effect on colors or style
• middle weight (about 1.2GB)

Summary

Use this LoRA when you want smoother, less chaotic background motion in Wan2.2 I2V videos.

It pairs especially well with more energetic or aggressive camera-movement models to keep the scene stable.