Keep what is already right
Regeneration often changes unrelated rooms. Canvas lets you focus on the specific wall segments that need correction.
AI floor plan editor
Open an AI result or existing image as a Canvas background. Trace the walls that need to change, undo freely, and preserve the correction as vector data.
Why this workflow works
Regeneration often changes unrelated rooms. Canvas lets you focus on the specific wall segments that need correction.
The background image remains intact while your vector wall layer can be added, removed, exported, or saved independently.
The JSON project format creates a path toward doors, windows, room labels, snapping, area calculation, and CAD export.
How it works
Keep the workflow simple: provide intent, review the AI draft, then make the changes that matter in Canvas.
Use the latest AI result, a sketch conversion, or a floor plan image from your generation history.
The image becomes a faded reference. Click start and end points to draw clean wall corrections over it.
Save the vector geometry as project JSON and export SVG when you need a portable result.
Questions, answered
Reference-image generation can attempt a local change, but model output may still affect unrelated areas. The Canvas layer is the more deterministic correction path.
The current beta does not destructively erase pixels. It adds a separate vector layer for clean correction and export.
The interface is responsive, but precise wall drawing is more comfortable on a tablet or desktop with a larger viewport.
Yes. The wall Canvas supports undoing the latest segment and clearing the current vector layer.