videoVideo to VideoVideo

Wan 2.2 14B Fun Control

Generate videos guided by pose, depth, and edge controls using Wan 2.2 Fun Control.

About

**About Control Video** You may need the following custom nodes to preprocess the control video: 1. comfyui_controlnet_aux 2. ComfyUI-DepthAnythingV2

You can find the control video we use here.

This model supports multiple control conditions such as Canny, Depth, Pose, and trajectory control.

Tutorial

Diffusion Model

LoRA

VAE

wan_2.1_vae.safetensors

Text Encoder

umt5_xxl_fp8_e4m3fn_scaled.safetensors

File save location

ComfyUI/
├─── models/
│   ├─── diffusion_models/
│   │   ├─── wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors
│   │   └─── wan2.2_fun_control_high_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

Note: About Wan2.2 Fun Control Wan2.2-Fun-Control is a 14B-parameter video generation model launched by Alibaba PAI, supporting multiple control conditions such as Canny, Depth, Pose and trajectory control.

It enables multi-resolution (512/768/1024) video prediction, trained with 81 frames and 16 FPS, and supports multilingual generation.

Using the Wan2.2 Lighting LoRA will result in the loss of video dynamics, but it will reduce the generation time. This template provides two workflows, and you can enable one as needed.

Parameters

Parameter	Type	Default	Description
`seed`	int	0	Random seed
`steps`	int	4	Number of sampling steps
`cfg`	float	1	CFG scale (guidance strength)
`sampler_name`	enum	euler	Sampler algorithm
`scheduler`	enum	simple	Noise scheduler
`filename_prefix`	string	video/wan2.2_fun_control	Output filename prefix

Usage

c3 comfyui run video_wan2_2_14B_fun_control \
  --prompt "your prompt here" \
  --output my_output \
  --steps 4 \
  --cfg 1

Required Models

Example Prompt

Prompt Example

"``` On a sunny summer day, there are marshmallow - like clouds, and the sunlight is bright and warm. A girl with white curly double - ponytails is wearing unique sunglasses, distinctive clothes and shoes. Her posture is natural and full of dynamic tension. The background is the scene of the Leaning Tower of Pisa in Italy, emphasizing the realistic contrast of details in reality. The whole picture is in a realistic 3D style, rich in details and with a relaxed atmosphere. She is dancing slowly, waving her hands. ```"

Ready to run this template?

Install the CLI and run this template on GPU in seconds.

pip install hypercli-cli && hyper comfyui run video_wan2_2_14B_fun_control

View Documentation