How long does it take to generate a video?

Using the 4-step Lightning LoRAs, generation typically takes around 70 seconds, whereas the 20-step standard high-quality pass can take 5-10 minutes.

Can I use this skill for Image-to-Video tasks?

This specific skill is optimized for Text-to-Video (T2V). For workflows involving image conditioning, you should use the wan-flf-video skill.

What is the benefit of the dual-model architecture in WAN 2.2?

It utilizes two specialized experts: the HighNoise model establishes motion and composition, while the LowNoise model focuses on refining textures and sharp details.

What hardware is required for WAN 2.2 video generation?

A GPU with 24GB VRAM (like an RTX 4090) is recommended for running dual FP8 models, though the skill provides optimization tips for lower VRAM environments.

WAN 2.2 Text-to-Video Workflows

Name: WAN 2.2 Text-to-Video Workflows
Author: artokun

byartokun

•

데이터 과학 및 ML

Builds and optimizes high-fidelity video generation pipelines using the WAN 2.2 14B Mixture-of-Experts architecture.

This skill provides expert guidance and automated configurations for creating WAN 2.2 Text-to-Video (T2V) workflows within ComfyUI. It specializes in the complex dual-model architecture, orchestrating the HighNoise model for structural composition and the LowNoise model for fine detail refinement. Users can leverage this skill to implement both 4-step Lightning workflows for rapid prototyping and 20-step standard pipelines for cinematic quality. It includes precise settings for ModelSamplingSD3 flow matching, VRAM management strategies for 14B parameter models, and integration for VACE control modules, ensuring professional-grade video output from text prompts.

주요 기능

01Integration with VACE modules for advanced reference and pose conditioning

02Support for ModelSamplingSD3 flow matching and EmptyHunyuanLatentVideo initialization

03Dual-pass HighNoise and LowNoise model orchestration for 14B MoE architecture

04Optimized VRAM management and resolution scaling up to 720p

05Automated configuration for Lightning LoRAs (4-step) and Standard (20-step) workflows

0611 GitHub stars

사용 사례

01Building rapid video prototyping pipelines using 4-step distilled lightning models

02Generating high-quality cinematic videos from descriptive text prompts

03Developing advanced ComfyUI workflows for Mixture-of-Experts video architectures

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add artokun/comfyui-mcp wan-t2v-video

For use in Claude.ai and ChatGPT

Download Skill