Configuring Flux nodes in ComfyUI

Step-by-step

1. Load the diffusion model using the Load Diffusion Model node
2. Set up two Clip Loader nodes to load multiple CLIP models simultaneously
3. Connect the conditioning to the Basic Guider node
4. Connect the Basic Guider to the Sampler Custom Advanced node
5. Configure the Schedulers (Sigmas) node for the specific Flux model type
6. Connect the Empty Latent Image node, setting the resolution to 1024x1024
7. Connect the Random Noise node directly into the Sampler Custom Advanced node
8. Load the VAE model using the VAE Loader and connect it to the VAE Decode node
9. Click Queue Prompt to begin the generation

Options

• Switch between pre-loaded models or manually uploaded model files
• Use Sampler Custom instead of Sampler Custom Advanced if negative conditioning is required
• Adjust batch size (e.g., set to 4) for multiple image generation

Watch out for

• Flux workflows do not use the traditional K Sampler node
• The standard Flux workflow shown lacks a negative conditioning input
• When switching to the Schnell model, you must reduce the steps to 4 to get it to work

Tips

• Use the RunComfy beta version to avoid spending hours downloading and uploading large model files
• Loading two clip loaders at once is recommended as it can improve the quality of the results
• Note that the K sampler selection is now handled within the sampler module itself rather than as a standalone node