Inspiration:We love choosing songs and music. We noticed the person on the aux never really reads the room. We wanted a DJ that listens to everyone at once.
What it does:Guests scan a QR code and share a vibe by text, photo, voice note, or favorite artist. Gemini picks every next song in any genre and explains why. Guests can scan their pulse so the DJ can feel the room’s energy. It never repeats a song and keeps heart rates private. An ElevenLabs voice announces each track.
How we built it:We broke the project into parts and followed a concrete plan. First we built the Gemini DJ brain that reads text, photos, and audio and picks songs. Then we built the vitals math that smooths noisy readings and compares each song to a baseline. Next came the Next.js host and guest apps, and the ElevenLabs voice. Finally we added the pulse scan with Presage, plus safety rules, tests, and hosting.
Challenges we ran into:Getting the Presage API working was our biggest struggle, because their API changed and needs video preprocessed by their own library. We also hit API rate limits and quotas, so we built a fallback chain of models. Keeping heart rates private while still guiding the music was tricky too. On top of that, we had to make the UI look polished under time pressure.
Accomplishments that we’re proud of:We built a full loop from guest input to an explained song. We designed a custom UI we’re proud of. The DJ handles any genre and never repeats. We also have 28 passing automated tests, including privacy checks.
What we learned:We learned how to connect several APIs and how each sponsor tool really works. We learned how to make an AI agent reliable with guardrails. We also learned how to design a clean, fun interface.
What’s next for DJ Sense:Next we want voice requests with Gemini Live and a continuous Presage pulse station. We also plan to save rooms in a database and add a shareable end-of-night recap.
Log in or sign up for Devpost to join the conversation.