Img2Img-ControlNetXL-Canny — Img2Img AI Model
Generate an SDXL image that follows an edge map: give it Canny outlines plus a prompt and the result keeps that structure. Sticks to the lines more tightly than the Union variant. Outputs up to 1024×1024. The control image can be up to 1024×1024.
What it does
ControlNet XL Canny generates a new image that follows the outlines of a Canny edge map while taking its style, mood and content from a text prompt.
It is a ControlNet adapter on SDXL: rather than letting the model invent a layout, you hand it the edge structure of a reference photo, a sketch or line art, and it fills in the rest.
When to use it
- Keeping a reference image's exact layout while changing everything else
- Turning a hand-drawn sketch or clean line art into a rendered image
- Restyling a scene in watercolor or anime without losing its geometry
- Alternate renderings of a building or product with structural lines kept
- Edge fidelity beyond what the general Union variant holds to
Details
- CategoryImg2Img
- Year2024
- LicenseApache-2.0
Compliance & Provenance
- ProviderOpen-source (xinsir) · Specialized
- EU AI Act RiskMinimal Risk
- Art. 50 TransparencyNot applicable
Inputs & Outputs
- Canny Edge MapInputimage
Canny edge map produced by the Annotator-CannyEdge node. White edges on black background. The control image is resized to match the output resolution. The control image can be up to 1024×1024.
- TextInputstring
Text prompt describing the desired output image. SDXL responds best to detailed English prompts. Include subject, style, lighting, and mood.
- ImageOutputimage
Generated image whose structure follows the canny edges, guided by the text prompt.
Parameters
- controlnet_conditioning_scalefloatdefault 1range 0–2
Edge guidance strength (0 = ignore, 2 = strict). 0.8–1.2 typical.
- stepsintegerdefault 25range 20–30
Diffusion steps (20–30).
- guidance_scalefloatdefault 5range 1–20
CFG scale (1–20, 5.0–7.5 typical for SDXL).
- seedintegerdefault -1range -1–2147483647
Random seed (-1 = random).
- widthintegerdefault 1024options: 512, 768, 1024
Output width (px). Edge map resized to match.
- heightintegerdefault 1024options: 512, 768, 1024
Output height (px). Edge map resized to match.
- negative_promptstringdefault advanced
Negative prompt — what to avoid. Empty for none.
Tags
Alternatives in Image Control
- ControlNet XL Union
Guided SDXL image generation from a control image + text prompt. One model, 6 modes: canny, depth, openpose, hed, normal, segment. Outputs up to 1024×1024.
- Depth Anything Annotator
Monocular depth estimation (Depth Anything V1 Large).