# WAN 2.2 Image-to-Video Models

High-performance Image-to-Video generation models for ComfyUI.

## Models Included

| File | Description | Size |
|------|-------------|------|
| `Wan2_2-I2V-A14B-HIGH_fp8_e4m3fn_scaled_KJ.safetensors` | HIGH noise I2V model | ~14GB |
| `Wan2_2-I2V-A14B-LOW_fp8_e4m3fn_scaled_KJ.safetensors` | LOW noise I2V model | ~1.3GB |
| `umt5_xxl_fp8_e4m3fn_scaled.safetensors` | Text encoder (UMT5-XXL) | ~6GB |
| `wan_2.1_vae.safetensors` | VAE decoder | ~243MB |

## Usage with RunPod

Set in RunPod endpoint settings:
- **Model**: `Aloukik21/wan22-i2v`
- **HF Token**: Your HuggingFace token

## Features
- I2V Original (20 steps)
- I2V Lightning (4-8 steps with LoRAs)
- First-Last Frame interpolation
- RIFE frame interpolation support