How to Install Wan2.2-Animate in ComfyUI (2026 Guide)

Q: What is Wan2.2 Animate?

Wan2.2 Animate is an AI tool for ComfyUI that animates characters in two ways: Animation mode makes a still image move by copying motion from a reference video, while Replacement mode swaps the person in a video with your character while keeping the original lighting and background.

Q: How do I fix out-of-memory errors with Wan2.2 Animate?

Reduce the Frame_load_cap (number of frames) and Force_rate (frame rate) in the Load Video node, and lower the output width and height. On low-VRAM GPUs, switch to the GGUF Q4_0 quantized model.

What is Wan2.2 Animate?

Quick answer: Wan2.2 Animate is an AI model for ComfyUI that can animate a still image by copying motion from a video, or swap a person in a video with your own character — preserving the original lighting and background.

Wan2.2 Animate is a tool that brings characters to life in video. You provide a character image and a short reference video, and it handles two distinct tasks:

Animation mode — Makes your character move, copy facial expressions, and follow body movements from the reference video.
Replacement mode — Swaps the original person in the video with your character while keeping the same lighting, colors, and background so the result looks natural.

In short, Wan2.2 Animate makes it easy to turn still images into moving, realistic animations — or smoothly replace people in existing video clips.

Wan2.2-Animate example — character animation and video replacement output

Download the Workflow

Download the ready-to-use Wan2.2 Animate workflow JSON file below. You can drag and drop it directly into your ComfyUI window after installation.

Install Wan2.2-Animate: Step-by-Step

Follow these instructions to install Wan2.2-Animate and set up the workflow in ComfyUI from scratch.

Step 1 — Update ComfyUI and All Nodes

Update ComfyUI

Go to your ComfyUI_windows_portable\update folder and double-click update_comfyui to update ComfyUI to the latest version.

Update All Nodes

Start the ComfyUI server. Open the Manager and click 'Update all' to ensure every custom node is on the latest version.

Step 2 — Import the Wan2.2 Animate Workflow

Download the JSON file above, then drag and drop it into the ComfyUI canvas.

⚠️ Warning: If you see red nodes after importing, some custom nodes are missing. Open the Manager, click "Install missing custom nodes", install everything required, then restart ComfyUI.

Step 3 — Download Required Models

Download the correct diffusion model for your GPU, plus the shared text encoder, VAE, and LoRA files.

Diffusion Models — choose one for your GPU

NVIDIA 4000 series (RTX 4070, 4080, 4090)Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors

NVIDIA 3000 series (RTX 3070, 3080, 3090)Wan2_2-Animate-14B_fp8_e5m2_scaled_KJ.safetensors

Low-end GPUs / 8 GB VRAM (GGUF quantized)Wan2.2-Animate-14B-Q4_0.gguf

You also need these shared files (all GPUs):

Text encoder: umt5-xxl-enc-bf16.safetensors
VAE: Wan2_1_VAE_bf16.safetensors
LoRA 1: WanAnimate_relight_lora_fp16.safetensors
LoRA 2: lightx2v_I2V_14B_480p_cfg_step_distill_rank256_bf16.safetensors

Step 4 — Place Models in the Correct Folders

Organize downloaded files exactly as shown below so ComfyUI can detect them automatically:

ComfyUI/
└── models/
    ├── diffusion_models/
    │   └── Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors
    │       (or e5m2 or GGUF version depending on your GPU)
    ├── text_encoders/
    │   └── umt5-xxl-enc-bf16.safetensors
    ├── vae/
    │   └── Wan2_1_VAE_bf16.safetensors
    └── loras/
        ├── WanAnimate_relight_lora_fp16.safetensors
        └── lightx2v_I2V_14B_480p_cfg_step_distill_rank256_bf16.safetensors

💡 Tip: Need example input files (character images and reference videos)? Download example input files from Google Drive →

Step-by-Step Workflow Setup

Once your models are in place and the workflow is imported, configure each node in order:

Load the Workflow

Drag and drop the downloaded JSON file into the ComfyUI canvas, or use the Load button in the interface.

Set the Diffusion Model

Select your downloaded diffusion model file in the "Load Diffusion Model" node.

Set the Text Encoder

Choose umt5-xxl-enc-bf16.safetensors in the "Load CLIP" node.

Set the VAE

Select Wan2_1_VAE_bf16.safetensors in the "Load VAE" node.

Set Both LoRA Files

Add both LoRA files in their respective Load LoRA nodes.

Add Your Video or Image

In the "Load Video" node, add your reference video or the character image you want to animate. Adjust Frame_load_cap and Force_rate if you get out-of-memory errors. Set width and height as needed.

Edit the Mask

Bypass the "Sampling Group" to adjust the mask using the Point Editor node. Once masking is set correctly, re-enable the Sampling Group.

Enter Your Prompt

Type a short description of the desired motion — for example: A girl is dancing

Click Queue Prompt

Click 'Queue Prompt' to run the workflow. The animation will be generated and saved to your ComfyUI output folder.

Example Run

Here is a real output generated with the workflow above — a character image animated using a reference dance video.

Wan2.2-Animate output — character animated from a reference video using the workflow above

Tips and Troubleshooting

If something goes wrong during setup or generation, one of these is almost always the cause:

Red nodes in the workflow

Fix: Open the ComfyUI Manager and click "Install missing custom nodes". Install all required nodes and restart ComfyUI.

Out-of-memory (OOM) error during generation

Fix: Reduce Frame_load_cap (fewer frames) and Force_rate (lower frame rate) in the Load Video node. Also lower the output width and height. On low-VRAM GPUs, switch to the GGUF Q4_0 model.

Very slow generation on a low-end GPU

Fix: Use the GGUF Q4_0 quantized model instead of the fp8 version. It runs faster and uses significantly less VRAM.

Poor animation quality or subject not tracked correctly

Fix: Use the Point Editor node to carefully select the subject area in the mask. Clear, high-quality input images and videos produce significantly better results.

Model not appearing in node dropdowns

Fix: Double-check that files are placed in exactly the correct sub-folder (diffusion_models, text_encoders, vae, loras). Refresh your browser after adding new files.

Still stuck? ComfyUI Troubleshooting Guide → covers a full list of errors and fixes across all workflow types.

Frequently Asked Questions

Wan2.2 Animate is an AI model for ComfyUI that animates characters in two ways: Animation mode makes a still image move by copying motion from a reference video, while Replacement mode swaps the person in a video with your character while keeping the original lighting and background.

Wan2.2 Animate works on NVIDIA 4000 series (fp8 e4m3fn model), NVIDIA 3000 series (fp8 e5m2 model), and lower-end GPUs (GGUF Q4_0 model). Choose the matching model file for your graphics card.

Place the diffusion model in ComfyUI/models/diffusion_models/, the text encoder in ComfyUI/models/text_encoders/, the VAE in ComfyUI/models/vae/, and the LoRA files in ComfyUI/models/loras/.

Red nodes mean one or more custom nodes required by the workflow are missing. Open the ComfyUI Manager, click "Install missing custom nodes", install all required nodes, then restart ComfyUI.

Reduce Frame_load_cap (number of frames) and Force_rate (frame rate) in the Load Video node, and lower the output width and height. On low-VRAM GPUs, switch to the GGUF Q4_0 quantized model.

Final Thoughts

Wan2.2 Animate makes it straightforward to turn images into compelling AI animations using ComfyUI. With the right model for your GPU, the free downloadable workflow, and these step-by-step instructions, anyone can get started quickly.

Experiment with different prompts, reference videos, and masking settings to discover what works best for your creative projects. The more precise your mask and the higher-quality your input, the better your results will be.