Control AI Generation with ControlNet
ControlNet adds structural guidance to the AI generation process, letting you control exactly how your output is composed. Feed in a pose skeleton and the AI generates a person in that exact position.
Try ControlNet NowWhat is ControlNet?
ControlNet works by processing a conditioning image (such as an edge map, depth map, pose skeleton, or normal map) alongside your text prompt. The conditioning image tells the AI where to place elements, how to structure the composition, and what spatial relationships to maintain, while the text prompt controls style, color, and subject matter. This separation of structure and style is incredibly powerful. You can take a photograph's pose, combine it with an anime style prompt, and get a perfectly posed anime character. Or extract the depth map from a landscape photo and regenerate it as a fantasy painting with the same spatial depth. Apatero supports multiple ControlNet modes including Canny (edge detection), Depth, Pose, Blur, and more.
Step-by-Step ControlNet Workflow
Follow these steps to get the best results with controlnet on Apatero
Choose Your Control Mode
Select the type of structural guidance you need. Canny extracts edges for precise outlines. Depth preserves spatial layout. Pose captures human body positions. Each mode is suited to different creative goals.
Upload or Generate a Control Image
Provide a reference image that contains the structure you want. Apatero automatically extracts the control signal (edges, depth, pose) from your image. You can also upload pre-processed control maps directly.
Write Your Style Prompt
Describe the visual style, subject, and mood you want. The ControlNet handles composition, so your prompt should focus entirely on aesthetics: colors, lighting, artistic style, and atmosphere.
Adjust Control Strength
Set how strictly the output follows the control image. High values (0.8-1.0) produce exact structural matches. Lower values (0.4-0.6) give the AI more freedom to interpret the guidance loosely.
Tips for Better ControlNet Results
Use these tips to get the most out of controlnet on Apatero
Use Canny mode for architectural scenes where you need precise straight lines and edges
Depth mode is best for landscapes and scenes where spatial relationships matter most
Pose mode works perfectly for character art, fashion illustrations, and action scenes
Combine ControlNet with a LoRA for consistent character appearance in controlled poses
Lower the control strength to 0.5-0.7 if the output looks too rigid or mechanical
Extract a pose from a stock photo and regenerate it in any art style for unique results
ControlNet FAQ
Common questions about controlnet on Apatero
What ControlNet modes does Apatero support?
Apatero supports Canny (edge detection), Depth (spatial depth maps), Pose (human body skeleton), Blur (soft structural guidance), and Normal maps. Each mode extracts different structural information from your reference image.
Do I need to create control maps manually?
No. Apatero automatically extracts control signals from any uploaded image. Just upload a regular photo and select the ControlNet mode. The platform handles edge detection, depth estimation, and pose extraction automatically.
Can I use multiple ControlNet modes at the same time?
Some workflows benefit from combining modes. For example, using Pose for character positioning and Depth for background layout simultaneously. Multi-ControlNet support depends on the base model and plan.
How is ControlNet different from img2img?
ControlNet extracts abstract structural information (edges, depth, pose) and uses it as a guide, giving complete freedom over colors and style. Img2img directly transforms the source pixels, so the original colors and textures influence the output more heavily.
What is the best control strength to start with?
Start at 0.75 for a good balance between structural accuracy and creative freedom. Increase to 0.9-1.0 for exact matches (like precise architectural lines), or decrease to 0.5 for looser, more artistic interpretations.