Enviza

Inspiration

Over the time my way of thinking and way of usage of tech has been changed due to my laziness i have got an idea where my room got messy didn't clean for a week and i just got 1hours to clean up so before i am out of time i wanted to make an organized room before i start to work so a pre organised image will be more easy and can be done in like 10 minutes

What it does

so upload around 20 images where a 3d model is made and the objects and things will be identified and now the user have the ability and access to replace there place which will make a organised room and it keeps the user to change the view every week

How we built it

We used Next.js and React to build the experience then added AI to help understand a room. Users can upload a room image or scan see which pieces of furniture Enviza recognizes drag an item to a new spot and preview how the room could look afterward. AI helps fill in the space left behind, and the before and after slider makes the change easy to compare.We also explored a multi-photo room scan feature. It takes several photos of a room and processes them into a 3D model using a separate reconstruction pipeline.

Challenges we ran into

Rooms are messy and personal. A chair might be easy to identify, but walls, built-in shelves, shadows, rugs, and overlapping furniture can confuse both AI and people. We had to make sure Enviza focused on things that can actually be moved instead of suggesting someone drag a wall across the room. The 3D scanning side was also challenging. Good room reconstruction depends on clear photos, enough overlap between shots, steady lighting, and nothing moving around in the room. We added checks and helpful error states so the experience feels honest instead of pretending every scan will work perfectly.

Accomplishments that we're proud of

We are proud that Enviza feels simple. The core experience is straightforward see your room, choose something to move, try a new layout, and decide whether you like it before doing the real work. We are also proud that we built more than a visual mockup. The project includes AI-powered object detection, an interactive editor, image healing for moved furniture, before-and-after comparison, and an experimental path toward full 3D room capture.

What we learned

We learned that people need to feel in control when AI is making suggestions about their own space. That is why Enviza asks for confirmation before changing anything and lets users compare the original room with the proposed version. We also learned that building for real spaces means planning for imperfect inputs. Photos can be blurry, lighting can change, and rooms are rarely as clean as a demo image. Designing around those imperfections was just as important as the AI itself.

What's next for Enviza

Next we want to make the 3D experience more complete better room scans, more accurate object placement, collision detection, and the ability to export a redesigned room for AR or sharing. Long term we want Enviza to become a helpful planning tool for moving, redecorating, buying furniture, or simply trying out a new idea so people can feel confident about a room change before lifting a single thing.

Built With

• celery
• codex
• colmap
• cuda
• docker
• fastapi
• gpt-image-2
• next.js
• openai-gpt-5.6
• postgresql
• python
• react
• redis
• s3/r2
• tailwind-css
• three.js
• typescript
• vercel