ComfyUI-Wan-Video-Integrated-KSampler

🐳 ComfyUI WanVideo Integrated KSampler

English | 简体中文

WanVideoIntegratedKSampler

This is an integrated ComfyUI WanVideo generation sampler node. Compared to using the official K-sampler, it eliminates the messy connections, supporting both text-to-video and image-to-video generation. It integrates dual-model sampling for high and low noise, single-frame generation, start/end frame generation, Motion Amplitude Enhancement, automatic GPU/RAM memory cleanup, batch generation, various attention optimizations, and more. Moms no longer have to worry about messy connections~~~~

If this project helps you, please give it a ⭐Star — it lets me know there are humans out there using it!

🏆 Features

🎬 Supported Generation Modes

• Text-to-Video: Generate videos from text prompts
• Image-to-Video: Generate videos from a single reference image
• Start/End Frame-to-Video: Generate videos from start and end frame images

⚡ Advanced Optimizations

• Dual-Stage Model Sampling: Integrated high and low noise dual-stage model sampling with configurable parameters for each stage
• Separate High/Low Noise Settings: Configure sampling steps and CFG values for high and low noise stages respectively
• Sage Attention Optimization: Integrated multiple attention optimization modes supporting memory-efficient computation
• FP16 Accumulation: Support Torch FP16 accumulation to improve VRAM utilization
• SD3 Sampling Integration: Integrated SD3 sampling algorithm, no additional nodes needed
• Motion Amplitude Enhancement: Integrated motion amplitude enhancement to improve video motion effects, reference project: PainterFLF2V

🖼️ Image Processing

• Integrated Prompt Input: Integrated prompt input box, no additional nodes needed
• Automatic Image Scaling: Automatically scales to target dimensions while maintaining aspect ratio
• Direct Video Frame Retrieval: Obtain video frame sequences directly, no additional decoding required
• Last Frame Retrieval: Obtain the last frame of generated video directly, no additional processing needed

🔧 Productivity Enhancer

• Batch Generation: Generate multiple videos in a single operation
• Automatic GPU Memory Cleanup: Automatic cleanup option for GPU/VRAM
• Automatic RAM Memory Cleanup: Automatic cleanup option for RAM
• CLIP Vision Enhancement: Support CLIP Vision model for enhanced perspective control
• Completion Sound Notification: Play audio alert upon completion

📦 Other

• Simple Version Provided: Includes only dual-model sampling integration for advanced custom use

🍧 Comparison Demo

🔄 Workflow Complexity Comparison

• ❌ Workflow without WanVideo Integrated KSampler (complex, many nodes, many connections, two samplers)
• ✅ Workflow with WanVideo Integrated KSampler (ultra-simple, single node, almost no connections, single sampler)

📦 Installation

Method 1: Via ComfyUI Manager (Recommended)

1. Open ComfyUI Manager in ComfyUI interface
2. Search for "ComfyUI-Wan-Video-Integrated-KSampler"
3. Click Install

Method 2: Manual Installation

1. Navigate to your ComfyUI custom nodes directory:

   cd /path/to/ComfyUI/custom_nodes
   

2. Clone the repository:

   git clone https://github.com/luguoli/ComfyUI-Wan-Video-Integrated-KSampler.git
   Or Gitee repository:
   git clone https://gitee.com/luguoli/ComfyUI-Wan-Video-Integrated-KSampler.git
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Restart ComfyUI

🚀 Usage

Workflow Example

Basic Text-to-Video Generation

1. Add the "🐳 WanVideo Integrated KSampler" node to your workflow
2. Connect the required inputs:
   • 🔥 High Noise Model
   • ❄️ Low Noise Model
   • 🟡 CLIP
   • 🎨 VAE
3. Input positive and negative prompts
4. Set relevant parameters:
   • Set video frame length, width, height
   • Configure sampling steps and CFG values for high/low noise
   • Set batch size, seed, etc.
5. Execute the workflow

Image-to-Video Generation

1. Add the node to the workflow
2. Connect required inputs:
   • 🔥 High Noise Model
   • ❄️ Low Noise Model
   • 🟡 CLIP
   • 🎨 VAE
3. Connect at least one reference image:
   • 🖼️ Start Image
   • 🖼️ End Image
4. Optionally connect 👁️ CLIP Vision for enhanced control
5. Input positive/negative prompts
6. Configure other parameters
7. Execute the workflow

⚠️ Notes

📝 Usage Requirements

• Text-to-Video Mode: No reference images needed, pure text prompts
• Image-to-Video Mode: At least one reference image required, start/end frames can be combined for precision control

🎛️ Parameter Recommendations

• Frame Length: Adjust based on GPU memory, start testing from 41 frames
• Resolution: Must be multiples of 8, start testing from 720x1280
• Batch Size: Choose between 1-10, adjust based on GPU memory
• Sampling Steps: Start testing from 4
• CFG Value: Default 1.0, recommended range 0.5-7.0
• FP16 Accumulation: Recommended to enable
• Sage Attention: Recommended to set to auto
• SD3 Shift: Recommend setting to 5
• Motion Amplitude: Default 1.0 (no enhancement), increase above 1.0 for stronger motion when using start/end frames

💾 Memory Management

• GPU Memory Cleanup: Enable enable_clean_gpu_memory to automatically clean VRAM before/after sampling
• CPU Memory Cleanup: Enable enable_clean_cpu_memory_after_finish to clean RAM after completion (includes file cache, processes, dynamic libraries),During continuous large-scale generation, always enable memory cleanup options to prevent memory overflow

🔊 Notification Function

• Sound Notification: Only supported on Windows systems

📝 Update Log

v1.0.1:

• Added translation script: Starting from ComfyUI v0.3.68, Chinese language files became invalid. Added an automatic translation script. Double-click [自动汉化节点.bat] and restart ComfyUI. Requires ComfyUI-DD-Translation plugin to be installed.
• Removed block swap settings: Latest ComfyUI version blocks block swapping, so this setting has been removed.

📞 Contact for Special Customization 📞

• Author: @luguoli（墙上的向日葵）
• Author Email: [email protected]

---

Made with ❤️ for the ComfyUI community