ComfyUI Integration

Custom nodes that expose WaveSpeed AI API features inside ComfyUI. Import the repository, configure your API key, and build image, video, or audio workflows without crafting REST calls manually.

Requirements

• WaveSpeed AI account with an active API key
• ComfyUI (portable build or source installation)
• Python environment that can install the package dependencies listed in requirements.txt

Installation

1. Open a terminal in your ComfyUI workspace and move into ComfyUI/custom_nodes.
2. Clone the repository:
   git clone https://github.com/WaveSpeedAI/wavespeed-comfyui.git
3. Install dependencies:
   • Windows portable ComfyUI: python -m pip install -r requirements.txt
   • Linux/macOS: pip install -r requirements.txt
4. Configure API key:
   • Either paste the key into the WaveSpeedAI Client node inside ComfyUI, or
   • Rename config.ini.tmp to config.ini and add the key so all nodes can read it automatically.
5. Restart ComfyUI. The WaveSpeed nodes appear in the node browser.

Recommended Workflow: Dynamic Nodes (WIP)

The dynamic node system replaces most individual model nodes. It lets you select any supported model at runtime and adjust parameters in one place.

• WaveSpeedAI Task Create [WIP]: Pick a model (Flux, Hunyuan, Wan2.1, Seedream, etc.) and edit the generated parameter list.
• WaveSpeedAI Task Submit [WIP]: Execute the prepared task, monitor status, and pull results.
• WaveSpeedAI Client: Provide credentials and optional defaults used by the other nodes.

Benefits

• Dynamic model selection without updating the repository for every new release.
• Parameter fields adapt to the chosen model, so you only see relevant inputs.
• Leaner graphs with fewer custom node types to maintain.

Basic Dynamic Workflow

1. Add WaveSpeedAI Client and set your API key if you did not use config.ini.
2. Drop WaveSpeedAI Task Create [WIP] into the graph.
3. Choose a model from the dropdown and configure the displayed parameters.
4. Connect its output to WaveSpeedAI Task Submit [WIP] and run the workflow.
5. Inspect the submit node output for task progress and generated assets.

Legacy Individual Model Nodes

Legacy nodes remain available while the dynamic approach is finalized. They cover Flux, Wan, Hunyuan, Kling, audio generation, LoRA helpers, and more. Expand the list inside ComfyUI to locate the specific node you need. We recommend migrating new workflows to the dynamic nodes, but you can continue running existing graphs unchanged.

Featured Workflows

• After cloning the repository, open the examples directory inside ComfyUI to view sample JSON workflows and annotated PNG graphs.
• Drag a sample JSON into ComfyUI to recreate the node graph and inspect how WaveSpeed nodes are wired.
• PNG screenshots embedded with workflow metadata can also be dropped into ComfyUI to load the corresponding graph.

LoRA Usage

1. Use nodes such as WaveSpeedAI Wan Loras, WaveSpeedAI Flux Loras, or WaveSpeedAI Flux SDXL Loras to load remote LoRA weights.
2. Provide a direct URL in lora_path (e.g., https://huggingface.co/WaveSpeedAi/WanLoras/resolve/main/wan_loras.safetensors).
3. Set lora_weight (for example, 0.5). Add multiple path/weight pairs to combine LoRAs.
4. Any publicly accessible URL works; local files cannot be uploaded directly through the API.

Working with image_url

1. Add WaveSpeedAI Upload Image to convert a local or binary input into an image_url.
2. Connect the node’s output to any downstream node that requires an image reference.
3. Review the example workflows for typical wiring patterns.

Support Resources

• Repository and issue tracker: https://github.com/WaveSpeedAI/wavespeed-comfyui