Create a room with Room Vision

Room Vision can analyze a written brief together with a floor plan or room photos, then propose editable room geometry.

Room Vision requires sign-in. A request can include up to eight JPG, PNG, WebP, or HEIC images, with a maximum source size of 25 MB each, plus a written brief of up to 600 characters. Images are compressed locally before analysis. The current allowance is three analyses in a rolling 24-hour period.

Prepare a strong input#

• Use a sharp floor plan that shows complete wall boundaries and readable dimensions.
• For a single room, include several angles when one photo cannot show the whole shape.
• Describe the target clearly—for example, “the main bedroom beside the balcony, 3 × 4 m.”
• Combine a plan with photos when you want both reliable geometry and visual context.

Review before applying#

Room Vision shows a preview before replacing room geometry. Compare the proposed walls, openings, shared boundaries, and dimension labels with the source. A safety check may block a result when shapes conflict, dimensions do not match, openings cannot attach to a wall, or the target room is ambiguous.

Applying an entire-home result can replace existing room geometry and remove content that no longer fits or remains attached to old walls. Read the confirmation carefully. The operation is grouped so one Undo can restore the previous state, but a named Library save before replacement is still the safest checkpoint.

AI analysis is not a site survey. Verify every wall, opening, and level before using the plan for purchasing, fabrication, or construction.