Achieving precise control over image creation in AI image generation cannot be done with just one click. It typically requires numerous generation attempts to produce a satisfactory image. However, the emergence of ControlNet has effectively addressed this challenge.

ControlNet is a conditional control generation model based on diffusion models (such as Stable Diffusion), first proposed by Lvmin Zhang and Maneesh Agrawala et al. in 2023 in the paper Adding Conditional Control to Text-to-Image Diffusion Models.

ControlNet models significantly enhance the controllability of image generation and the ability to reproduce details by introducing multimodal input conditions, such as edge detection maps, depth maps, and pose keypoints.

These conditioning constraints make image generation more controllable, allowing multiple ControlNet models to be used simultaneously during the drawing process for better results.

Before ControlNet, we could only rely on the model to generate images repeatedly until we were satisfied with the results, which involved a lot of randomness.

With the advent of ControlNet, we can control image generation by introducing additional conditions. For example, we can use a simple sketch to guide the image generation process, producing images that closely align with our sketch.

In this example, we will guide you through installing and using ControlNet models in ComfyUI, and complete a sketch-controlled image generation example.

The workflows for other types of ControlNet V1.1 models are similar to this example. You only need to select the appropriate model and upload the corresponding reference image based on your needs.

​
ControlNet Image Preprocessing Information

Different types of ControlNet models typically require different types of reference images:

Image source: ComfyUI ControlNet aux

Since the current Comfy Core nodes do not include all types of preprocessors, in the actual examples in this documentation, we will provide pre-processed images. However, in practical use, you may need to use custom nodes to preprocess images to meet the requirements of different ControlNet models. Below are some relevant custom nodes:

​
ComfyUI ControlNet Workflow Example Explanation

​
1. ControlNet Workflow Assets

Please download the workflow image below and drag it into ComfyUI to load the workflow:

Images with workflow JSON in their metadata can be directly dragged into ComfyUI or loaded using the menu Workflows -> Open (ctrl+o). This image already includes download links for the corresponding models, and dragging it into ComfyUI will automatically prompt for downloads.

Please download the image below, which we will use as input:

​
2. Manual Model Installation

If your network cannot successfully complete the automatic download of the corresponding models, please try manually downloading the models below and placing them in the specified directories:

ComfyUI/
├── models/
│   ├── checkpoints/
│   │   └── dreamCreationVirtual3DECommerce_v10.safetensors
│   ├── vae/
│   │   └── vae-ft-mse-840000-ema-pruned.safetensors
│   └── controlnet/
│       └── control_v11p_sd15_scribble_fp16.safetensors

In this example, you could also use the VAE model embedded in dreamCreationVirtual3DECommerce_v10.safetensors, but we’re following the model author’s recommendation to use a separate VAE model.

​
3. Step-by-Step Workflow Execution

  1. Ensure that Load Checkpoint can load dreamCreationVirtual3DECommerce_v10.safetensors
  2. Ensure that Load VAE can load vae-ft-mse-840000-ema-pruned.safetensors
  3. Click Upload in the Load Image node to upload the input image provided earlier
  4. Ensure that Load ControlNet can load control_v11p_sd15_scribble_fp16.safetensors
  5. Click the Queue button or use the shortcut Ctrl(cmd) + Enter to execute the image generation

​
Load ControlNet Node Explanation

Models located in ComfyUI\models\controlnet will be detected by ComfyUI and can be loaded through this node.

​
Apply ControlNet Node Explanation

This node accepts the ControlNet model loaded by load controlnet and generates corresponding control conditions based on the input image.

Input Types

Parameter NameFunction
positivePositive conditioning
negativeNegative conditioning
control_netThe ControlNet model to be applied
imagePreprocessed image used as reference for ControlNet application
vaeVAE model input
strengthStrength of ControlNet application; higher values increase ControlNet’s influence on the generated image
start_percentDetermines when to start applying ControlNet as a percentage; e.g., 0.2 means ControlNet guidance begins when 20% of diffusion is complete
end_percentDetermines when to stop applying ControlNet as a percentage; e.g., 0.8 means ControlNet guidance stops when 80% of diffusion is complete

Output Types

Parameter NameFunction
positivePositive conditioning data processed by ControlNet
negativeNegative conditioning data processed by ControlNet

You can use chain connections to apply multiple ControlNet models, as shown in the image below. You can also refer to the Mixing ControlNet Models guide to learn more about combining multiple ControlNet models.

You might see the Apply ControlNet(Old) node in some early workflows, which is an early version of the ControlNet node. It is currently deprecated and not visible by default in search and node lists. To enable it, go to Settings —> comfy —> Node and enable the Show deprecated nodes in search option. However, it’s recommended to use the new node.

​
Start Your Exploration

  1. Try creating similar sketches, or even draw your own, and use ControlNet models to generate images to experience the benefits of ControlNet.
  2. Adjust the Control Strength parameter in the Apply ControlNet node to control the influence of the ControlNet model on the generated image.
  3. Visit the ControlNet-v1-1_fp16_safetensors repository to download other types of ControlNet models and try using them to generate images.

Was this page helpful?

Suggest edits
Multiple LoRAsPose ControlNet