princepainter/ComfyUI-PainterI2V

What it solves

This project addresses the "slow-motion" effect often found in Wan2.2 image-to-video (I2V) generation, particularly when using 4-step LoRAs like lightx2v. It increases the motion amplitude and makes camera movement prompts more responsive.

How it works

It replaces the standard WanImageToVideo node in ComfyUI with a specialized PainterI2V node. The system employs three core technical approaches:

• Brightness-Protected Motion Scaling: It separates the luminance mean before scaling motion vectors to prevent brightness shifts during movement enhancement.
• Zero Latent Initialization: It preserves the temporal dependency chain necessary for 4-step LoRAs.
• Reference Frame Enhancement: It uses reference_latents to ensure the subject remains consistent while allowing for freer motion.

Who it’s for

Users of ComfyUI and the Wan2.2 model who want more dynamic, faster-paced movement in their AI-generated videos without sacrificing subject consistency.

Highlights

• Motion Boost: Increases motion amplitude by 15% to 50% via a motion_amplitude parameter.
• Plug & Play: Fully compatible with original Wan2.2 workflows.
• Scene-Specific Tuning: Provides recommended settings for different movement types, from smooth walking (1.10‐1.20) to extreme sports (1.5).
• Optimized for I2V: Specifically designed for single-frame image-to-video workflows.