ControlNet: Image guidance ControlNet (canny)

Keep the layout of a picture you have and let the prompt decide the rest, using a map of its features.

8 stepsBring a picture of your own

The same tutorial, three subjects

Step by step

Explanation detail

  1. Model & style

    1 · Model

    Select Deliberate for this example.The studio's ControlNet mode requires a supported Stable Diffusion 1.5 model. Check compatibility and worker availability before generating.Set to 1. The “Set this for me” button selects Deliberate and sets CLIP skip to 1. Models can need different CLIP skip settings; using the wrong one can give poor results.

  2. Source image

    2 · Source mode

    Choose ControlNet under Source mode.The worker creates a control map from your picture, such as outlines or a pose. This guides the new image alongside your prompt. The result can still differ from the original structure.
  3. Source image

    3 · Source image

    ControlNet uses features from an existing image to guide a new one.Open , choose a Source mode to show the upload control, and attach your image. The tutorial cannot attach a file for you.The example uses an image at 512×512 pixels. Match those dimensions for a similar comparison.
  4. Source image

    4 · Control type

    Choose canny for this example's edge map.Other types capture different features. Depth estimates how far objects are from the camera; pose detects body positions; surface normals describe which way surfaces face.Select the information you want to preserve and inspect the map if the result is unexpected.
  5. Source image

    5 · Denoising strength

    Use 1. In this ControlNet workflow, the source supplies a map that guides generation from noise.Use to adjust the map's influence. Denoising strength does not specify how closely outlines or poses must match.

  6. Prompt

    6 · Prompt

    Describe the subject and appearance you want, including medium, colours, and lighting. The map and prompt both influence the result.Prompt:
    ink and watercolour illustration of a woman in her 30s with a sleek silver bob haircut, round copper goggles resting up in her silver hair, a patched teal canvas coat, a small brass compass on a cord around her neck, at the wooden railing of a zeppelin airship's open gondola deck, the canvas envelope curving overhead, a sea of clouds below, loose brush strokes, soft washes of colour, visible paper texture
    Negative prompt:
    photograph, 3d render, lowres, blurry, deformed hands, extra fingers, watermark, text
    An edge map does not prescribe a photographic style. A negative prompt can discourage an appearance, but cannot ensure its removal.

  7. Size

    7 · Dimensions and sampling

    Set to 512×512. Match the source's aspect ratio to avoid stretching its structure.Recorded sampling settings: k_euler_a, 30 steps, guidance 7.5, seed 1234567890.

  8. 8 · Generation

    Press Generate. The worker creates the map from your attached image and uses it to guide generation.The job needs a worker that supports your model and control type. Follow the queue estimate. This run took about 11 seconds.

Related tutorial

Try generating an image with transparency so you can place its subject on another background.