Structure control generates a completely new image that keeps the layout of your photo or sketch. Shapes and composition come from the reference; content, style and light come from your description.
How it works in Pixelift
- Reference: you upload a photo or sketch (JPG, PNG or WEBP, up to 20 MB).
- Mode: you choose depth or edges.
- Description: you write what should be generated. The description is required, and the model expands it before generating.
- Result: you get one PNG image.
Generation settings are fixed. There is no strength slider and no pose mode; you steer the result with the mode and your description.
Two modes
Depth (FLUX Depth Pro)
The model reads a depth map from the reference, meaning what is near and what is far, and builds the new image in the same space.
Best for: interiors, architecture, landscapes and scenes with a clear foreground and background.
Edges (FLUX Canny Pro)
The model extracts an edge map from the reference with the Canny algorithm and follows the outlines, so shapes stay in place.
Best for: sketches, graphics, logos, products and objects with clear outlines.
Which mode to choose
| Goal | Mode |
|---|---|
| The same room with a new design | Depth |
| A building in another style or season | Depth |
| A sketch turned into a finished image | Edges |
| A product in a new material and color | Edges |
Not sure? Run both modes on the same photo and compare.
How to write the description
- Describe style, materials, light and mood, e.g. "Scandinavian interior, light wood, morning light".
- Do not describe the layout; it comes from the reference.
- Result too strict or too loose? Change the mode or the description, as the strength of the reference cannot be set.
Preparing the reference
- For edges: a sharp photo or sketch with clear lines and good contrast.
- For depth: a photo with readable perspective; flat images give weak depth maps.
- Crop to what should be preserved.
Limitations
- Only the structure is kept, not the content of the reference.
- There is no pose control, no segmentation and no combining of several modes.
- Complex scenes need a refined description.
Structure control vs other tools
| Need | Tool |
|---|---|
| A new image with the reference layout | Structure Control |
| A sketch turned into an image from a description | Sketch to Image |
| A ready-made style applied to a photo | Style Transfer |
| A free variation of a photo | AI Reimagine |
| Changing part of an image | Inpainting |
Cost and access
Structure Control requires an active Starter plan or higher. The current price per generation is shown on the tool page and on the button; credits are returned if generation fails. You can enlarge the finished result with the upscaler.