Best Preprocessing
Models and preprocessing abstractions that transform an input image into an intermediate guidance artifact for downstream workflows. Useful for ControlNet conditioning, structure-preserving edits, pose transfer, segmentation-guided generation, and other source-informed image pipelines.
Best rated
ControlNet Preprocess Tile prepares an image for tile-style ControlNet conditioning, emphasizing local detail and patch-level consistency. It is useful for upscaling-adjacent workflows, texture refinement, and preserving fine local structure when a model needs more detailed guidance from the source image.
Featured Models
Top-performing models in this category, recommended by our community and performance benchmarks.
ControlNet Preprocess Shuffle creates a shuffled representation of the source image for ControlNet conditioning. It is useful when the goal is to borrow palette, local texture, and visual character from a source image without rigidly preserving exact edges, contours, or pose.
ControlNet Preprocess SoftEdge extracts smoother, more organic edge structure from an input image for ControlNet conditioning. It is useful when the goal is to preserve contours and layout without the harsh binary boundaries of Canny, especially for painterly, naturalistic, or softer illustration workflows.
ControlNet Preprocess Lineart Anime extracts anime-style line structure from an input image for ControlNet conditioning. It is best for manga, cel-shaded, and stylized character workflows where cleaner region boundaries and illustration-style contours matter more than photoreal edge detail.
ControlNet Preprocess Depth estimates a depth map from an input image for ControlNet conditioning. It is useful when the goal is to preserve scene layout, foreground-background separation, camera perspective, and overall 3D spatial structure while changing style or content.
ControlNet Preprocess MLSD extracts major straight line segments from an input image for ControlNet conditioning. It is especially effective for buildings, interiors, product shots, and other compositions where rigid linear structure matters more than texture or organic detail.
ControlNet Preprocess Canny converts an input image into a high-contrast Canny edge map for ControlNet conditioning. It is best suited to workflows that need clear object boundaries, strong silhouette retention, and tightly constrained structure during guided image generation or editing.
ControlNet Preprocess Scribble converts an input image into a simplified scribble-style guide for ControlNet conditioning. It is best for workflows that want loose structural guidance, rough composition transfer, or sketch-based reinterpretation instead of strict edge preservation.
ControlNet Preprocess Lineart extracts simplified line art from an input image for ControlNet conditioning. It is useful when the goal is to preserve contours, composition, and object shapes while translating the result into illustration, concept art, or other stylized outputs.
ControlNet Preprocess Seg converts an input image into a semantic segmentation map for ControlNet conditioning. It is best for preserving broad scene organization such as sky, ground, buildings, clothing, vegetation, and other region-level structure while allowing major stylistic changes.
ControlNet Preprocess NormalBae predicts a surface normal map from an input image for ControlNet conditioning. It is well suited to workflows that need stronger preservation of 3D form, object volume, and orientation cues beyond what a flat edge map or line map can provide.
ControlNet Preprocess OpenPose extracts a human pose skeleton from an input image for ControlNet conditioning. It is best for character and figure workflows where body posture, gesture, limb placement, and overall pose need to stay consistent across generations.