Scribble Detection

Extract organic contour maps and hand-drawn line sketches from source photos for flexible ControlNet spatial conditioning

Extract scribble map

White magnolia flower and branch contours on black, extracted by Scribble Detection using HED with Safe Mode.

How Scribble Detection works

Convert photographic source images into expressive outline conditioning maps in three straightforward actions.

A magnolia photograph connected to Scribble Detection in Fuser, with its extracted contour map.

Upload source photograph

Provide a well-lit photograph with clear separation between the subject and background.

Scribble Detection in Fuser showing the HED algorithm selection and its magnolia contour output.

Select detection algorithm

Choose PiDi for clean ink-like edges or HED for smooth volumetric shading contours.

Scribble Detection in Fuser with Safe Mode enabled and its extracted magnolia outline.

Configure edge safety

Toggle Safe Mode to quantize gradients and eliminate source color bleed in downstream renders.

What Scribble Detection is good at

High-contrast contour extraction powered by dual edge-detection algorithms and edge-quantization controls.

White lounge-chair outline on black, extracted by Scribble Detection using PiDiNet.

Dual edge-detection engines

Switch between PiDi for crisp, low-noise ink outlines and HED for soft, charcoal-like gradient maps that capture subtle volumetric depth.

White watering-can contour on black, extracted using Scribble Detection with HED and Safe Mode.

Quantized Safe Mode processing

Enable Safe Mode to quantize continuous edge gradients into discrete stepped strokes, preventing source color bleed and ghosting in downstream generative renders.

White contours of three gourds on linen against black, extracted by Scribble Detection using HED.

Volumetric shape preservation

Preserve essential structural envelopes and volumetric boundaries from complex photographs, giving generative pipelines layout anchors without locking in surface textures.

White outline of an airborne dancer on black, extracted by Scribble Detection using PiDiNet.

Flexible spatial conditioning

Isolate organic sketch boundaries that retain subject framing and dynamic poses, providing high-level spatial guidance for generative re-renderings.

Made with Scribble Detection

Explore how organic sketch maps capture photographic composition while leaving textures open for creative reimagining.

White contour of a woman in a flowing gown on black, extracted by Scribble Detection using HED.

Loose fabric drape and dynamic silhouette extraction

White bicycle frame and wheel outlines on black, extracted by Scribble Detection using PiDiNet.

Sharp mechanical boundary extraction with PiDiNet

White windsurfer, sail and horizon contours on black, extracted by Scribble Detection using HED.

Atmospheric motion and framing guidance

White curved sculpture and plinth contours on black, extracted by Scribble Detection using HED.

Organic sculptural contour capture with HED

White headphone outlines on black, extracted by Scribble Detection using PiDiNet with Safe Mode.

Quantized outline definition for product restyling

What people build with Scribble Detection

Creative directors, visual artists, and designers rely on scribble maps to preserve composition while reinventing visual style.

Fashion lookbook restyling

Convert runway model photography into organic scribble maps to explore novel fabric textures, colorways, and seasonal styling while locking silhouette poses.

Product concept iteration

Transform physical mockups or prototype photos into clean contour guides to experiment with exotic finishes, metals, and composite materials.

Architectural visualization

Extract structural envelopes from site photos to render alternate weather conditions, evening illumination, or avant-garde cladding materials.

Cinematic storyboarding

Turn production stills and location scouts into expressive scribble maps for rapid scene restyling, lighting redesigns, and mood exploration.

Brand campaign adaptations

Capture the core geometry of brand assets to rapidly generate themed seasonal campaign imagery without distorting hero product placement.

Scribble Detection specs

Plans and credits

Versions: Holistically-Nested Edge Detection, PiDi
Inputs: Image
Output: image
Cost: 5 credits per run

Frequently Asked Questions

PiDi (Pixel Difference Networks) extracts crisp, clean outlines with minimal background noise, making it the best default choice for structural layout control. HED (Holistically-Nested Edge Detection) captures multi-scale, softer gradient maps resembling charcoal sketches, preserving subtle volumetric forms and facial expressions at the cost of slight background noise.

Safe Mode quantizes smooth grayscale edge gradients into stepped, distinct outlines resembling felt-tip marker strokes. This pre-quantization prevents ghosting artifacts where colors and textures from the source photograph bleed into downstream generations. Disable Safe Mode when you need maximum fine detail, as quantization can lead downstream models to hallucinate textures across gradient gaps.

Well-lit photographs with strong subject-to-background contrast yield the cleanest scribble maps. Avoid uploading existing black-and-white sketches or clean line art, as re-processing them causes double-edge detection artifacts and muddy outlines.

Reach for Scribble Detection when you want to restyle a composition while allowing downstream generative models full freedom to reinvent textures, lighting, and materials. Switch to Canny or Lineart detectors when you require rigid, millimeter-precise edge tracking for architectural blueprints or typography.

Try Scribble Detection on Fuser

Extract organic contour maps and hand-drawn line sketches from source photos for flexible ControlNet spatial conditioning

Extract scribble map

Scribble Detection docs All models

FLUX.3 Image

Black Forest Labs

Gemini Image

Google DeepMind

GPT Image

OpenAI

Seedream

ByteDance (Seed team)

FLUX.2