An installable web app where a shopper uploads a photo of their own room, picks an item from the catalog, places it in true perspective and gets a photorealistic rendering of that item in that room, with an admin area for the catalog, per-use-case model choice and a small content module.
What ports as-is
- The browser canvas renderer (bilinear warp with Newton iteration for the inverse mapping)
- Corner-handle and body-drag placement
- Floor and wall modes
- Rotation and drop shadows
- Anchor-based calibration for true-to-scale size
- The .NET 10 catalog and AI API
- A vision room-analysis step returning vanishing point and a wall-suitability verdict with a graceful default
- A room-analysis cache
- A placement quality-assurance loop that scores each generated image and feeds rejection reasons into the next attempt's prompt
- An admin-configurable model registry
- A PWA shell
- CI pipelines and an end-to-end suite
What we build for you
- The catalog itself (records, texture images, real-world dimensions) and the default room-size calibration for the category
- Brand skin and a voice guide for UI copy
- Sample room photos
- Model selection and keys
- Social cards
- Domain
- Slot and database
The perspective math, the QA loop and the admin area are configuration.
