Configure ControlNet

Configures one ControlNet entry — model, control image, and conditioning strength — for use with the ControlNet Pipeline node.

Category: ModularDiffusion/Conditioning

TL;DR

• Pick provider first; the rest of the parameters are dynamic and regenerate per provider.
• One node = one ControlNet. To stack multiple, drop multiple nodes and connect each control_net output to the ControlNet Pipeline node's control_nets list input.
• Supported providers: Flux, Qwen, Stable Diffusion, Z-Image etc.

Typical workflow position

Load Image → [Configure ControlNet] → ControlNet Pipeline → Generate Media Latents

Node preview

Inputs

Name	Type	Required	Notes
control_image	ImageArtifact / ImageUrlArtifact	Yes	The control signal (canny edges, depth map, pose, etc. — depends on the ControlNet model).
controlnet_conditioning_scale	float (0.0–1.0)	No	Influence of this ControlNet, default 1.0.
controlnet_model	HF repo picker	Yes	The ControlNet model to use; choices update per provider.

Outputs

Name	Type	Notes
control_net	control_net (dict)	Entry to feed into the ControlNet Pipeline node.

Provider / model behavior

The model dropdown and extra parameters change per provider. Notable extras:

• Flux — control_mode (canny, tile, depth, blur, pose, gray, low_quality), control_guidance_start, control_guidance_end.
• Qwen / Stable Diffusion / Z-Image — provider-specific model repos and conditioning knobs.

Tips & pitfalls

• Provider must match the base pipeline. A Flux ControlNet plugged into an SDXL pipeline will be rejected by the ControlNet Pipeline node's validator.
• Preprocess control images upstream. Run your detector (Canny, depth estimator, pose) in an earlier node and feed the result directly — this node expects a ready-to-use control image.
• controlnet_conditioning_scale is per-ControlNet. When stacking, each node has its own weight.

See also

• ControlNet Pipeline — required downstream consumer.
• Workflow template: workflows/templates/ControlnetText2Image.py.