comfyui_controlnet_aux

AIO_Preprocessor

Runs one selected ControlNet auxiliary preprocessor on an image at the requested resolution, supplying that preprocessor’s declared defaults for its other controls. Selecting “none” passes the image through unchanged.

ControlNet Preprocessors · 1 output

AnimalPosePreprocessor

Detects animal pose landmarks with the selected bounding-box detector and pose estimator. It returns both a rendered animal-pose map and structured pose-keypoint data.

ControlNet Preprocessors/Faces And Poses Estimators · 2 outputs

AnimeFace_SemSegPreprocessor

Runs semantic segmentation tuned for anime faces and returns the color-coded segmentation map. It can optionally remove the background with the Anime Background Remover path before segmentation.

ControlNet Preprocessors/Semantic Segmentation · 2 outputs

AnimeLineArtPreprocessor

Extracts anime-style line art from an image with the pack’s pretrained LineartAnime detector at the chosen resolution.

ControlNet Preprocessors/Line Extractors · 1 output

AnyLineArtPreprocessor_aux

Combines MTEED/AnyLine edges with a selected standard, realistic, anime, or manga line-art detector. It can limit line-art intensity and remove small connected objects before merging the two maps.

ControlNet Preprocessors/Line Extractors · 1 output

BAE-NormalMapPreprocessor

Estimates a surface-normal map from an image with the pretrained NormalBae detector at the chosen resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

BinaryPreprocessor

Converts an image into a binary line or silhouette map using a configurable intensity threshold and processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

CannyEdgePreprocessor

Extracts a Canny edge map from an image with configurable low and high thresholds and processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

ColorPreprocessor

Reduces an image to a simplified color-palette control map with the pack’s Color detector at the chosen resolution.

ControlNet Preprocessors/T2IAdapter Only · 1 output

ControlNetAuxSimpleAddText

Draws a supplied text label in green at the upper-left of the first image. It is used by the all-preprocessors diagnostic output to identify each generated map.

ControlNet Preprocessors · 1 output

ControlNetPreprocessorSelector

Provides one supported auxiliary-preprocessor identifier through a typed selector output so another node can choose the processing method.

ControlNet Preprocessors · 1 input · 0 parameters · 1 output

DensePosePreprocessor

Runs DensePose on an image using the selected model and color-map mode, returning a dense human-surface correspondence image at the requested resolution.

ControlNet Preprocessors/Faces And Poses Estimators · 1 output

DepthAnythingPreprocessor

Estimates a relative depth map with a selectable Depth Anything ViT checkpoint and returns it at the chosen processing resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

DepthAnythingV2Preprocessor

Estimates a depth map with a selected Depth Anything V2 checkpoint, including relative and metric indoor/outdoor variants, at the requested resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

DiffusionEdge_Preprocessor

Extracts learned edge maps with an indoor, urban, or natural DiffusionEdge checkpoint. Patch batch size trades VRAM for throughput while resolution controls preprocessing size.

ControlNet Preprocessors/Line Extractors · 1 output

DSINE-NormalMapPreprocessor

Estimates a surface-normal map with DSINE using configurable field of view, refinement iterations, and processing resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

DWPreprocessor

Detects human body, hand, and/or face poses with selectable DWPose detection and estimation models. It returns a rendered pose image and structured pose-keypoint data, with optional stick scaling for Xinsir ControlNet.

ControlNet Preprocessors/Faces And Poses Estimators · 2 outputs

ExecuteAllControlNetPreprocessors

Builds and executes a graph containing every supported auxiliary preprocessor, labels each result with its preprocessor name, and combines the results into one image batch for comparison.

ControlNet Preprocessors · 1 output

FacialPartColoringFromPoseKps

Converts face landmarks in pose-keypoint data into a color-coded facial-part image. Skin, eyes, nose, lips, and inner mouth can be drawn as colored points or filled polygons.

ControlNet Preprocessors/Pose Keypoint Postprocess · 1 output

FakeScribblePreprocessor

Creates a scribble-like control map from HED soft edges by applying the pack’s fake-scribble processing at the requested resolution.

ControlNet Preprocessors/Line Extractors · 1 output

HEDPreprocessor

Extracts a soft edge map from an image with the pretrained HED detector at the selected resolution.

ControlNet Preprocessors/Line Extractors · 1 output

HintImageEnchance

Resizes a ControlNet hint map to generation dimensions using direct resize, outer fit with padding, or inner fit with cropping. Its high-quality path preserves palette maps, binary regions, one-pixel edges, and inpainting alpha masks.

ControlNet Preprocessors · 1 input · 2 parameters · 1 output

ImageGenResolutionFromImage

Reads an image tensor and returns its width and height as generation-resolution integers.

ControlNet Preprocessors · 1 input · 0 parameters · 2 outputs

ImageGenResolutionFromLatent

Reads a latent tensor’s spatial dimensions and returns the corresponding image width and height using ComfyUI’s eight-times latent scale.

ControlNet Preprocessors · 1 input · 0 parameters · 2 outputs

ImageIntensityDetector

Builds an intensity-based recolor control map from an image with optional gamma correction and processing resolution.

ControlNet Preprocessors/Recolor · 1 output

ImageLuminanceDetector

Builds a luminance-based recolor control map from an image with optional gamma correction and processing resolution.

ControlNet Preprocessors/Recolor · 1 output

InpaintPreprocessor

Combines an image and mask into the special image hint used by ControlNet inpainting. An option switches the masked area to black pixels for Xinsir-compatible ControlNet models.

ControlNet Preprocessors/Others · 1 output

LeReS-DepthMapPreprocessor

Estimates a LeReS depth map with controls for removing near objects, removing background, boosted processing, and output resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

LineArtPreprocessor

Extracts realistic line art with the pretrained Lineart detector, with an optional coarse mode and configurable resolution.

ControlNet Preprocessors/Line Extractors · 1 output

LineartStandardPreprocessor

Produces conventional line art with Gaussian-smoothing and intensity-threshold controls at the requested resolution.

ControlNet Preprocessors/Line Extractors · 1 output

M-LSDPreprocessor

Detects straight line segments with M-LSD using configurable score and distance thresholds and processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

Manga2Anime_LineArt_Preprocessor

Extracts manga-oriented line art with the pretrained Manga Line detector at the selected resolution for manga-to-anime conditioning.

ControlNet Preprocessors/Line Extractors · 1 output

MaskOptFlow

Restricts an optical-flow tensor to a supplied mask, resizing the mask to the flow field and multiplying both flow and its colorized preview. It returns the masked flow and preview image.

ControlNet Preprocessors/Optical Flow · 2 outputs

MediaPipe-FaceMeshPreprocessor

Detects and renders MediaPipe facial meshes with controls for maximum faces, minimum detection confidence, and processing resolution.

ControlNet Preprocessors/Faces And Poses Estimators · 1 output

MeshGraphormer-DepthMapPreprocessor

Uses MeshGraphormer to generate hand depth maps and a corresponding inpainting mask. Controls select mask derivation, bounding-box padding, mask expansion, seed, detection thresholds, and resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 2 outputs

MeshGraphormer+ImpactDetector-DepthMapPreprocessor

Uses an external Impact Pack bounding-box detector to locate regions, then runs MeshGraphormer on each crop to build hand depth maps and inpainting masks. Detector threshold, dilation, crop factor, drop size, and MeshGraphormer mask settings are exposed.

ControlNet Preprocessors/Normal And Depth Estimators · 2 outputs

Metric_DepthAnythingV2Preprocessor

Compact description unavailable for this node.

Metric3D-DepthMapPreprocessor

Estimates metric depth with a selected Metric3D ViT backbone and camera focal lengths, returning the depth channel at the chosen resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

Metric3D-NormalMapPreprocessor

Estimates surface normals with a selected Metric3D ViT backbone and camera focal lengths, returning the normal channel at the chosen resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

MiDaS-DepthMapPreprocessor

Estimates a MiDaS depth map with configurable normal-map parameter, background threshold, and processing resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

MiDaS-NormalMapPreprocessor

Uses MiDaS depth inference to produce a surface-normal map with configurable normal strength, background threshold, and processing resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

OneFormer-ADE20K-SemSegPreprocessor

Runs OneFormer semantic segmentation with the ADE20K label set and returns the color-coded segmentation map at the chosen resolution.

ControlNet Preprocessors/Semantic Segmentation · 1 output

OneFormer-COCO-SemSegPreprocessor

Runs OneFormer semantic segmentation with the COCO label set and returns the color-coded segmentation map at the chosen resolution.

ControlNet Preprocessors/Semantic Segmentation · 1 output

OpenposePreprocessor

Detects body, hand, and/or face pose landmarks in an image. It outputs both a rendered OpenPose image and structured pose-keypoint data, with controls for resolution and which landmark groups to include.

ControlNet Preprocessors/Faces And Poses Estimators · 2 outputs

PiDiNetPreprocessor

Extracts PiDiNet soft edges with an optional safe-mode quantization and configurable processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

PixelPerfectResolution

Calculates a ControlNet preprocessor resolution from the original image, target generation size, and resize mode so the hint scale matches inner-fit, outer-fit, or direct-resize behavior.

ControlNet Preprocessors · 1 input · 2 parameters · 1 output

PyraCannyPreprocessor

Extracts multi-scale PyraCanny edges with configurable low and high thresholds and processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

RenderAnimalKps

Renders structured AP-10K animal pose-keypoint data into an animal skeleton image at the stored canvas dimensions.

ControlNet Preprocessors/Pose Keypoint Postprocess · 1 input · 0 parameters · 1 output

RenderPeopleKps

Renders structured human pose-keypoint data into an OpenPose-style image, with independent switches for body, hands, and face landmarks.

ControlNet Preprocessors/Pose Keypoint Postprocess · 1 input · 3 parameters · 1 output

SAMPreprocessor

Runs the Segment Anything automatic mask generator on an image and returns its segmentation visualization at the requested resolution.

ControlNet Preprocessors/Others · 1 output

SavePoseKpsAsJsonFile

Writes structured human or animal pose-keypoint data to a numbered JSON file in ComfyUI’s output directory.

ControlNet Preprocessors/Pose Keypoint Postprocess · 1 input · 1 parameter · 0 outputs

Scribble_PiDiNet_Preprocessor

Creates a scribble control map with PiDiNet, with optional safe-mode processing and a configurable resolution.

ControlNet Preprocessors/Line Extractors · 1 output

Scribble_XDoG_Preprocessor

Creates a scribble-style line map with an extended Difference-of-Gaussians detector using a configurable threshold and resolution.

ControlNet Preprocessors/Line Extractors · 1 output

ScribblePreprocessor

Converts an image into a basic scribble control map at the requested processing resolution.

ControlNet Preprocessors/Line Extractors · 1 output

SemSegPreprocessor

Legacy alias for the UniFormer semantic segmenter. It returns a color-coded semantic-segmentation map at the selected resolution.

ControlNet Preprocessors/Semantic Segmentation · 1 output

ShufflePreprocessor

Produces a seeded content-shuffled image control map at the chosen resolution, preserving coarse color and texture while disrupting spatial arrangement.

ControlNet Preprocessors/T2IAdapter Only · 1 output

TEED_Preprocessor

Compact description unavailable for this node.

ControlNet Preprocessors/Line Extractors · 1 output

TEEDPreprocessor

Extracts soft edges with the TEED detector, with a safe-steps control for quantizing or stabilizing the edge response and a configurable resolution.

ControlNet Preprocessors/Line Extractors · 1 output

TilePreprocessor

Runs tile preprocessing by shrinking an image according to the selected iteration count and rebuilding it with pyramid upsampling. It returns the resulting image for downstream tile ControlNet conditioning.

ControlNet Preprocessors/Tile · 1 output

TTPlanet_TileGF_Preprocessor

Creates a TTPlanet tile-conditioning map with scale, blur, guided-filter radius, and epsilon controls.

ControlNet Preprocessors/Tile · 1 output

TTPlanet_TileSimple_Preprocessor

Creates a simplified TTPlanet tile-conditioning map with scale-factor and blur-strength controls.

ControlNet Preprocessors/Tile · 1 output

UniFormer-SemSegPreprocessor

Runs UniFormer semantic segmentation and returns a color-coded scene-label map at the chosen resolution.

ControlNet Preprocessors/Semantic Segmentation · 1 output

Unimatch_OptFlowPreprocessor

Estimates optical flow between consecutive frames in an image batch with a selected UniMatch/GMFlow checkpoint. It can compute backward or bidirectional flow and returns both the flow tensor and a colorized preview sequence.

ControlNet Preprocessors/Optical Flow · 2 outputs

UpperBodyTrackingFromPoseKps

Converts human pose-keypoint sequences into per-frame bounding-box tracks for selected people and upper-body parts. It returns InstanceDiffusion-style tracking data plus a prompt fragment naming each tracked part and person ID.

ControlNet Preprocessors/Pose Keypoint Postprocess · 1 input · 1 parameter · 2 outputs

Zoe_DepthAnythingPreprocessor

Estimates metric depth with the Zoe Depth Anything indoor or outdoor checkpoint and returns the depth map at the requested resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output

Zoe-DepthMapPreprocessor

Estimates a monocular depth map with the pretrained Zoe detector at the selected processing resolution.

ControlNet Preprocessors/Normal And Depth Estimators · 1 output