Inspiration
We wanted to make interior design feel like a live creative conversation, where an AI decorator can see a room, understand the vibe, and instantly suggest a transformed version.
What it does
Decorate Live Agent guides a live room scan, remembers the room and the user’s style preferences, finds visual inspiration, and generates a redesigned room render in the UI.
How we built it
We built it with FastAPI, Google ADK, Gemini Live, Vertex AI Search, Firestore, Cloud Storage, and an image-generation pipeline grounded in room snapshots plus inspiration references.
Challenges we ran into
Our biggest challenges were handling real-time multimodal state, keeping the agent grounded in what it actually saw, and making the generation flow reliable end to end.
Accomplishments that we're proud of
We’re proud that the experience feels like a real design session, with live voice interaction, memory across steps, inspiration retrieval, and a final render that stays anchored to the actual room.
What we learned
We learned that great multimodal experiences depend as much on orchestration, memory, and prompt discipline as they do on the underlying models.
What's next for Decorate Live Agent
Next, we want to improve inspiration quality, support user-provided reference images, add iterative revision loops, and make the redesigns even more controllable and consistent.
Log in or sign up for Devpost to join the conversation.