HomeAR

• 

  Three different design layouts come by default

• 

  Editing a particular image using text is supported

• 

  Three different design layouts come by default

• 

  Screen Capture

• 

  Editing a particular image using text is supported

Inspiration

I was inspired by how difficult it can be to visualize new designs in a room without actually remodeling it. Most design apps I’ve seen either feel complex or rely on prebuilt 3D models. I wanted something simple — capture a photo of my space, give a prompt, and instantly see how it could look.

What it does

HomeAR lets me capture a photo of my room and instantly reimagine it with AI-powered design variations. From “modern minimal” to “luxury wood paneling,” I can swipe through multiple design styles, refine them with text prompts, and even chat with each generated image to make iterative edits. It feels like having a personal designer available on demand.

How we built it

I built a React Native app with Expo so it works across devices. The app captures photos and sends them to a Flask backend. On the backend, I integrated image generation API (Flux model) on my my mac to create variations conditioned on the captured photo. I return those results to the app, display them in a smooth carousel, and added a chat-with-image feature to make the editing process conversational and interactive.

Challenges we ran into

Making sure the AI generated fresh results instead of cached ones took some debugging. Expo Go threw runtime errors when dealing with camera and layout, which I had to carefully debug and refactor. On the backend, I had to handle multiple requests at once without overwriting image outputs, so I implemented thread locks and unique file handling.

Accomplishments that we're proud of

I built a full pipeline in the hackathon timeframe — capture → AI variations → chat edits → updated results in real-time. I designed a smooth UI where swiping, prompting, and chatting with images feels intuitive. I proved that it’s possible to combine AI image generation with mobile AR-style visualization in a lightweight way.

What we learned

I learned how to manage caching and async image processing in mobile apps. I also deepened my understanding of working with Image Models on system and via APIs and handling Expo’s quirks with the camera and rendering. Most importantly, I learned how much thoughtful UX (like variation labels and chat buttons) can improve the whole experience.

What's next for HomeAR

Next, I want to extend the app into true AR by letting users project their AI-designed rooms back into their physical space using ARKit/ARCore. I also want to explore connecting HomeAR with furniture and home décor marketplaces so users can shop directly for the styles and items they see in their generated images.

Built With

• ai
• ar
• augmented-reality
• flux
• github
• gpu
• ml
• python
• react
• reactnative