Controlnet Node
The Controlnet node generates outputs through conditioning on an input image. It has 4 subtypes: Flux Pro, Flux Dev, Stable Diffusion XL (legacy) and Stable Diffusion 1.5 (Legacy)
Flux Pro Controlnet
The Flux Pro Controlnet node is powered by Flux Pro technology. It generates images conditioned on images using a specific preprocessor type.
Inputs
The image that will be used for conditioning. This image will be passed through the preprocessor before being sent to the controlnet.
The prompt that will be used to generate the image.
Settings
Add up to 3 colors using the color palette to select.
Switch on this setting to allow the AI to add details to your prompt to improve generations. This helps image quality, but can occasionally lead to worse prompt adherence.
Guidance is a scale from 0 to 100 that measures how much the output image should adhere to the prompt. Increasing guidance to values that are too high result in a lack of coherence — the images follow the prompt, but they won’t look quite right.
Increasing the number of steps increases the number of iterations the AI takes to generate your image. Increasing steps make generations take longer, but can sometimes lead to more coherent images.
The canny edge detection preprocessor creates an edge map of the image, and uses this edge map to condition the output image. The output image will follow the edge map of the input image, but use the prompt to generate.
The depth map preprocessor creates a depth map of the image, and uses this depth map to condition the output image. The output image will follow the depth map of the input image, but use the prompt to generate.
Outputs
The output image that is conditioned by the processed base image, and the prompt.
Flux Dev Controlnet
The Flux Dev Controlnet node is powered by Flux Dev technology. It generates images conditioned on images using a specific preprocessor type. You can view the effects of a preprocessor on an image by using a “Controlnet Preprocess” node.
Inputs
The image that will be used for conditioning. This image will be passed through the preprocessor before being sent to the controlnet.
The prompt that will be used to generate the image.
Settings
Add up to 3 colors using the color palette to select.
Switch on this setting to allow the AI to add details to your prompt to improve generations. This helps image quality, but can occasionally lead to worse prompt adherence.
Guidance is a scale from 0 to 100 that measures how much the output image should adhere to the prompt. Increasing guidance to values that are too high result in a lack of coherence — the images follow the prompt, but they won’t look quite right.
Increasing the number of steps increases the number of iterations the AI takes to generate your image. Increasing steps make generations take longer, but can sometimes lead to more coherent images.
The canny edge detection preprocessor creates an edge map of the image, and uses this edge map to condition the output image. The output image will follow the edge map of the input image, but use the prompt to generate.
Conditioning Scale: A number between 0-1, that increases the strength of the conditioning.
High Threshold: A number between 0 and 255, that is a threshold for strong edges. Any pixel value higher than the threshold is taken as a “strong edge”.
Low Threshold: A number between 0 and 255, that is a threshold for weak edges. Any pixel value lower than the threshold is not taken as an edge
The depth map preprocessor creates a depth map of the image, and uses this depth map to condition the output image. The output image will follow the depth map of the input image, but use the prompt to generate.
Conditioning Scale: A number between 0-1, that increases the strength of the conditioning.
The OpenPose preprocessor creates a stick figure of the subject in the image, reflecting their pose.
Conditioning Scale: A number between 0-1, that increases the strength of the conditioning. Hand: If turned on, the pose will also take the position of the fingers of the hand into account.
Outputs
The output image that is conditioned by the processed base image, and the prompt.