What is ControlNet?

ControlNet is an open-source neural network architecture (introduced in 2023) that conditions an image-generation model on a structural reference — a pose skeleton, depth map, edge map, segmentation map, or sketch — alongside the text prompt. It is the technical foundation for 'I want this exact pose with this product on this background' control in modern AI image generation. Stable Diffusion, Flux, and most open-source image models support ControlNet conditioning. For commercial AI UGC, ControlNet-style control is what separates 'pretty AI image' from 'campaign-ready asset': you can lock the product's exact position, the persona's exact pose, and the camera angle, then iterate on lighting, scene, and styling. ppl.studio's persona consistency and product placement run on a ControlNet-class control layer under the hood — though most users interact through high-level scene presets rather than the raw conditioning maps.