ComfyUI Workflow: ControlNet Canny (FLUX)
This is the official Canny ControlNet example from the ComfyUI FLUX examples. Canny edge detection extracts the structural lines of a source image (the silhouette, the major edges, the layout), and FLUX uses those edges as a generation guide. The prompt fills in the style, lighting, and details, while the structure stays locked.
Canny is the most common ControlNet preprocessor and the right choice when you want to keep the composition of a reference but change everything else. Architecture renders that keep the building shape but change the materials, product shots that keep the layout but change the color, and character concept art that keeps the pose but restyle the character all fit this workflow.
Key benefits
- Canny preprocessor handles the edge map automatically, no external tool required.
- FLUX.1 base model produces high quality, photoreal outputs on consumer hardware with the right quant.
- Locked composition with full style freedom in the prompt.
- Official ComfyUI example, maintained alongside the core.
How it works
- Download the workflow PNG from the linked GitHub URL and drag it into ComfyUI.
- Get the model files: FLUX.1 fp8 checkpoint goes in
models/checkpoints, Canny controlnet weights go in models/controlnet. Both are linked in the FLUX examples README.
- Load your source image into the LoadImage node. The Canny preprocessor runs on it and extracts the edge map.
- Write the prompt describing what you want the output to look like. Style, lighting, materials, mood.
- Queue the prompt. Generation time depends on the FLUX checkpoint size and your GPU.
Canny strength controls how strictly the output follows the source edges. High strength keeps the silhouette tight. Low strength lets the model take more creative liberty with the composition.