ControlNet is an advanced AI image control system that gives you precise control over pose, depth, and spatial relationships in your AI-generated images. Unlike standard text-to-image generation where results can be unpredictable, ControlNet lets you lock exact poses (DWPose) and spatial relationships (ZoeDepth) for reproducible, professional results. Only Nano Banana Pro offers this level of control - Midjourney and DALL-E do not support ControlNet.

DWPose is a pose control model that extracts skeletal structure from reference images and locks exact body positions, limb angles, and posture. Use DWPose when you need consistent poses across multiple generations (character reference sheets, product photography). ZoeDepth is a depth estimation model that creates 3D spatial understanding and locks volume, depth, and camera-to-body distance. Use ZoeDepth to prevent the 'flattening' effect and maintain realistic spatial relationships. You can use both together for maximum control.

Weight values (0.0-1.0) control how strongly ControlNet influences the generation. Higher values = stricter adherence to your reference. For DWPose, use 1.0 for exact pose lock (no deviation), 0.7-0.9 for strong control with minor adjustments, 0.5-0.7 for moderate flexibility. For ZoeDepth, use 0.8-1.0 for strong depth lock (recommended for product/portrait), 0.5-0.7 for some artistic freedom. Values below 0.5 provide weak control and are not recommended for consistency.

For text-to-image generation without a reference, ControlNet parameters provide structured guidance but work best when combined with a reference image. When you upload a reference image, DWPose and ZoeDepth analyze the pose and depth from that image, then lock those characteristics in the generation. Without a reference, ControlNet uses your parameter constraints as guidance, but results may be less precise. For professional-grade consistency, always upload a reference image when using ControlNet.

Yes, you can use ControlNet parameters for text-to-image generation without uploading a reference image. The structured JSON prompt provides detailed guidance on pose, depth, constraints, and negative prompts. However, for the most precise control, we recommend uploading a reference image so DWPose and ZoeDepth can analyze exact pose and depth characteristics. The parameter builder tool works for both scenarios with or without reference.

ControlNet generation with Nano Banana Pro costs 20 credits per generation. This is the same rate as all Nano Banana Pro advanced features. Credit pricing starts at $4.99 for 550 credits, with bulk discounts up to 3000 credits for $19.99. Each generation includes both pose control (DWPose) and depth control (ZoeDepth) when configured in your parameters.

For DWPose pose control, we recommend 1.0 for exact pose lock when consistency is critical (character sheets, product catalogs). Use 0.8-0.9 for strong control with minor adjustments. For ZoeDepth depth control, we recommend 0.8 for most use cases (portraits, products, architecture) - this provides strong depth lock while maintaining natural appearance. Use 0.9-1.0 when exact spatial relationships are essential. Lower values (0.5-0.7) provide more creative freedom but less control.

Use the Negative Prompt Builder to toggle forbidden elements that cause AI beautification. Key presets include: 'beautification filters', 'skin smoothing', 'plastic skin', 'airbrushed texture', 'body proportion averaging', 'reduced curves', 'slimmed torso', 'flattened glutes', and 'aesthetic proportion correction'. These 19 preset categories prevent AI from applying beauty standards and preserve authentic body characteristics. You can also add custom forbidden elements for specific needs.

Yes, the parameter builder tool includes preset management. Click 'Save as Preset' to store your configuration locally with a custom name. Use 'Load Preset' to recall saved configurations. We also include built-in presets for common scenarios (Character Sheet, Product Photo, Fashion Editorial) to help you get started. Saved presets are stored in your browser's localStorage and persist between sessions.

Nano Banana Pro is the only web-based AI image tool that offers ControlNet integration with DWPose and ZoeDepth. Midjourney and DALL-E do not support ControlNet - they cannot lock exact poses or depth, leading to inconsistent results. Stable Diffusion has ControlNet but requires technical knowledge, local installation, and manual JSON coding. Nano Banana Pro eliminates technical barriers with a visual interface, real-time JSON preview, and educational tooltips, making professional-grade control accessible to everyone.