Momento

Prompting the vision_agent in the web adk
Output as seen in the web adk
Image input field in the client-facing frontend
Relevant output in the client-facing frontend

Inspiration

We were inspired by a friend struggling to find the perfect song for his Instagram post. We realized that every photo carries its own emotional signature, and matching it with music could create a seamless, expressive experience. Momento was born from the idea of turning a simple image into a personalized soundtrack.

What it does

Momento accepts a photo as input and uses an AI agent to analyze its emotional tone. The agent generates a structured estimate of the song’s features—a musical vector representing mood, energy, danceability, acousticness, instrumentalness, and tempo. It then searches Spotify playlists matching the identified mood, fetches songs, converts them into feature vectors, and computes cosine similarity between each song’s vector and the estimated vector from the image. The song with the closest match is recommended, along with the artist and reasoning behind the selection.

How we built it

We built Momento using the Google Agent Development Kit to create an intelligent music-matching agent, Python for backend logic, and the Spotify API to fetch playlists, track metadata, and audio features. We implemented a cosine similarity calculation to rank songs based on feature similarity. A React frontend with Tailwind CSS provides an interactive, visually appealing interface for users to upload images and explore their personalized musical matches.

Challenges we ran into

Integrating a sophisticated AI agent with a React frontend presented a unique challenge. We also had to learn new tools, navigate the Spotify API (including handling missing or deprecated features), and ensure smooth real-time communication between the backend and frontend.

Accomplishments that we're proud of

We successfully implemented a fully functioning AI agent capable of analyzing images, estimating audio feature vectors, and recommending songs using cosine similarity matching. On top of that, we created a polished demo interface where users can instantly see and hear the musical counterpart to their photo.

What we learned

We gained hands-on experience with AI agent design, vector-based feature matching, and building a complete pipeline connecting image input to music recommendation. We also deepened our understanding of React, Python, API integration, and designing a user-friendly interface for complex AI processes.

What's next for Momento

Our next step is to fully and elegantly integrate the underlying agent with an engaging, intuitive interface. We aim to provide a seamless experience where users can explore a wide variety of moods, playlists, and personalized recommendations—all powered by AI and ML techniques that match images to audio features.