Skip to main content

​
About Wan2.1-Fun-InP

Wan-Fun InP is an open-source video generation model released by Alibaba, part of the Wan2.1-Fun series, focusing on generating videos from images with first and last frame control. Key features:
  • First and last frame control: Supports inputting both first and last frame images to generate transitional video between them, enhancing video coherence and creative freedom. Compared to earlier community versions, Alibaba’s official model produces more stable and significantly higher quality results.
  • Multi-resolution support: Supports generating videos at 512x512, 768x768, 1024x1024 and other resolutions to accommodate different scenario requirements.
Model versions:
  • 1.3B Lightweight: Suitable for local deployment and quick inference with lower VRAM requirements
  • 14B High-performance: Model size reaches 32GB+, offering better results but requiring higher VRAM
Below are the relevant model weights and code repositories:
Currently, ComfyUI natively supports the Wan2.1 Fun InP model. Before starting this tutorial, please update your ComfyUI to ensure your version is after this commit.
Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Wan2.1 Fun InP Workflow

​
1. Download Workflow

Download the image below and drag it into ComfyUI to load the workflow: Wan2.1 Fun InP Workflow

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “Wan 2.1 Inpainting” in Template Library

Start Image

Starting frame for the video generation. Download and use this image, or replace with your own.

End Image

Ending frame for the video generation. Download and use this image, or replace with your own.

​
2. Manual Model Installation

If automatic model downloading is ineffective, please download the models manually and save them to the corresponding folders. All models involved in this guide can be found at Wan_2.1_ComfyUI_repackaged and Wan2.1-Fun. Diffusion Models : Choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements:

Diffusion Model: Wan2.1 Fun InP 1.3B

models/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors

Diffusion Model: Wan2.1 Fun InP 14B

models/diffusion_models/Wan2.1-Fun-14B-InP.safetensors (rename after download)
Text Encoders : Choose one of the following models (fp16 precision has a larger size and higher performance requirements):

Text Encoder: umt5_xxl_fp16

models/text_encoders/umt5_xxl_fp16.safetensors

Text Encoder: umt5_xxl_fp8

models/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
VAE

VAE: wan_2.1_vae

models/vae/wan_2.1_vae.safetensors
CLIP Vision

CLIP Vision: clip_vision_h

models/clip_vision/clip_vision_h.safetensors
File storage location:

​
3. Complete the Workflow Step by Step

ComfyUI Wan2.1 Fun InP Video Generation Workflow Diagram
  1. Ensure the Load Diffusion Model node has loaded wan2.1_fun_inp_1.3B_bf16.safetensors
  2. Ensure the Load CLIP node has loaded umt5_xxl_fp8_e4m3fn_scaled.safetensors
  3. Ensure the Load VAE node has loaded wan_2.1_vae.safetensors
  4. Ensure the Load CLIP Vision node has loaded clip_vision_h.safetensors
  5. Upload the starting frame to the Load Image node (renamed to Start_image)
  6. Upload the ending frame to the second Load Image node
  7. (Optional) Modify the prompt (both English and Chinese are supported)
  8. (Optional) Adjust the video size in WanFunInpaintToVideo, avoiding overly large dimensions
  9. Click the Run button or use the shortcut Ctrl(cmd) + Enter to execute video generation

​
4. Workflow Notes

Please make sure to use the correct model, as wan2.1_fun_inp_1.3B_bf16.safetensors and wan2.1_fun_control_1.3B_bf16.safetensors are stored in the same folder and have very similar names. Ensure you are using the right model.
  • When using Wan Fun InP, you may need to frequently modify prompts to ensure the accuracy of the corresponding scene transitions.