Wan2.2 Fun InP ComfyUI Workflow Complete Usage Guide, Official + Community Versions (Kijai, GGUF)

Tutorial Overview

This tutorial will comprehensively introduce various implementation methods and usage of the Wan2.2 Fun InP first-last frame video generation model in ComfyUI. Wan2.2 Fun InP is a professional first-last frame control video generation model launched by Alibaba Cloud, supporting input of first and last frame images to generate intermediate transition videos, bringing stronger creative control to creators.

Versions and Content Covered in This Tutorial

Completed Versions:

• ✅ ComfyUI Official Native Version - Complete workflow provided by ComfyOrg official
• ✅ Wan2.2 Fun InP 14B First-Last Frame Version - High-quality first-last frame control video generation

Versions in Preparation:

• 🔄 Kijai WanVideoWrapper Version - Community-developed convenient wrapper
• 🔄 GGUF Quantized Version - Optimized version for low-configuration devices

Model Technical Features

Wan2.2 Fun InP is based on the Wan2.2 architecture and has been specifically optimized for first-last frame control video generation, with the following core features:

Core Advantages:

• First-Last Frame Control: Supports input of first and last frame images to generate intermediate transition videos, enhancing video coherence and creative freedom
• High-Quality Video Generation: Based on the Wan2.2 architecture, outputting cinema-level quality videos
• Multi-Resolution Support: Supports generating videos at resolutions such as 512×512, 768×768, 1024×1024, adapting to different scenario requirements

The Wan2.2 Fun InP series models are based on the Apache2.0 open source license, supporting commercial use. The Apache2.0 license allows you to freely use, modify and distribute these models, including commercial purposes, as long as you retain the original copyright notice and license text.

Wan2.2 Fun InP Open Source Model Version Overview

Official Learning Resources

Related Code Repositories

• VideoX-Fun GitHub Repository - Official complete implementation code
• Wan2.2 Fun Official Documentation - Detailed model description and usage guide

Wan2.2 Fun InP ComfyUI Official Native Version Workflow Usage Guide

Version Description

The ComfyUI official native version is provided by the ComfyOrg team, using repackaged model files to ensure optimal compatibility with ComfyUI. This version supports both standard mode and Lightx2v 4-step LoRA acceleration mode.

Performance Comparison Test

Below are the test results using RTX4090D 24GB VRAM GPU for 640*640 resolution and 81-frame length:

Since using accelerated LoRA provides significant speed improvement, although there is some loss in dynamics, it is more user-friendly for low VRAM users. Therefore, in the two provided workflows, the accelerated LoRA version is enabled by default. If you need to enable the other workflow, select it and use Ctrl+B to enable it.

1. Wan2.2 Fun InP First-Last Frame Video Generation ComfyUI Workflow

Workflow Acquisition Method

Use the materials below as first and last frames

Model File Download

Diffusion Model

• wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors
• wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors

Lightning LoRA (Optional, for acceleration)

• wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors
• wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors

VAE

• wan_2.1_vae.safetensors

Text Encoder

• umt5_xxl_fp8_e4m3fn_scaled.safetensors

ComfyUI/
├───📂 models/
│   ├───📂 diffusion_models/
│   │   ├─── wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors
│   │   └─── wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors
│   ├───📂 loras/
│   │   ├─── wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors
│   │   └─── wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors
│   ├───📂 text_encoders/
│   │   └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors 
│   └───📂 vae/
│       └── wan_2.1_vae.safetensors

Detailed Operation Steps

1. High noise model and LoRA loading

• Ensure the Load Diffusion Model node loads the wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors model
• Ensure the LoraLoaderModelOnly node loads the wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors

2. Low noise model and LoRA loading

• Ensure the Load Diffusion Model node loads the wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors model
• Ensure the LoraLoaderModelOnly node loads the wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors

3. Ensure the Load CLIP node loads the umt5_xxl_fp8_e4m3fn_scaled.safetensors model
4. Ensure the Load VAE node loads the wan_2.1_vae.safetensors model
5. First and last frame image upload, upload first and last frame image materials respectively
6. Enter prompts in the Prompt group
7. WanFunInpaintToVideo node size and video length adjustment
   • Adjust the width and height dimensions, default is 640, we have set a smaller size and you can modify it as needed
   • Adjust length, this is the total number of video frames. The current workflow fps is 16. If you need to generate a 5-second video, you should set 5*16 = 80
8. Click the Run button, or use the shortcut Ctrl(cmd) + Enter to execute video generation

Wan2.2 Fun InP Kijai WanVideoWrapper ComfyUI Workflow

This part of the tutorial will introduce the convenient method using Kijai/ComfyUI-WanVideoWrapper.

Related model repository: https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled

Wan2.2 Fun InP GGUF Quantized Version ComfyUI Workflow

The GGUF version is suitable for users with limited VRAM, providing the following resources:

QuantStack/Wan2.2-Fun-A14B-InP-GGUF

Related Custom Nodes: City96/ComfyUI-GGUF