ComfyUI Integration

ComfyUI Integration

ComfyUI custom nodes for WaveSpeedAI. Run Flux, Hunyuan, Kling, Wan, and 70+ AI models directly in ComfyUI for text-to-image, image-to-video, and audio generation — no REST API coding required.

Installation

  1. Open ComfyUI Manager
  2. Search for “WaveSpeed”
  3. Click Install
  4. Restart ComfyUI

Manual Installation

cd ComfyUI/custom_nodes
git clone https://github.com/WaveSpeedAI/wavespeed-comfyui.git

Install dependencies:

Windows:

python -m pip install -r requirements.txt

Linux / macOS:

pip install -r requirements.txt

Setup

Add your API key in one of two ways:

  • Paste directly into the WaveSpeedAI Client node, or
  • Rename config.ini.tmp to config.ini and add your key there

Restart ComfyUI — WaveSpeed nodes will appear in the node browser.

Supported Models

CategoryModels
Text-to-ImageFlux (Dev, Schnell, Pro), Seedream, Hunyuan, Stable Diffusion
Image-to-VideoKling, Wan, Hunyuan, Vidu, LTX
Text-to-VideoWan, Hunyuan, LTX
Image UpscalingReal-ESRGAN
Audio & TTSMMAudio, Dia TTS
LoRAFlux LoRA, Wan LoRA, SDXL LoRA

The dynamic node system lets you select any model at runtime without updating the repository:

  • WaveSpeedAI Task Create [WIP] — Select model and configure parameters
  • WaveSpeedAI Task Submit [WIP] — Execute task and retrieve results
  • WaveSpeedAI Client — Provide API credentials

Parameters automatically adapt to the selected model, keeping your workflow clean.

Helper Nodes

NodePurpose
WaveSpeedAI Upload ImageConvert local images to URLs for API input
WaveSpeedAI Media Images To ListOrganize multiple image URLs into an array

LoRA Support

Load LoRA models from any public URL. Supports 6+ LoRAs simultaneously in a single generation.

  1. Add a LoRA node (Wan Loras, Flux Loras, or Flux SDXL Loras)
  2. Enter the URL in lora_path
  3. Set lora_weight (e.g., 0.5)
  4. Combine multiple LoRAs by adding more path/weight pairs

Example Workflows

The repository includes ready-to-use workflow examples:

Dynamic Node Examples:

  • Nano Banana image generation
  • Seedream V4 generation
  • Seedream V4 sequential processing
  • Qwen LoRA application
  • Advanced task chaining (image-to-image -> image-to-video)

Showcase Workflows:

  • Seedance video generation
  • Hunyuan Ref2Video
  • Wan 2.1 Image-to-Video

Open the examples directory or drag any workflow JSON/PNG into ComfyUI to load it.

Legacy Nodes

70+ individual model nodes remain available for backward compatibility. New workflows should use the dynamic nodes, but existing graphs will continue to work unchanged.

© 2025 WaveSpeedAI. All rights reserved.