Momento

Inspiration

As an event lead, it is so difficult to come up with ideas for an event. Hence, Momento was created to help with event planning as well as creating a long-lasting memory from events. We especially thought of events like hackathons, university club events, and social mixers, where small teams are trying to plan something memorable together, often without a clear starting point.

What it does

Memento turns event planning into a visual, collaborative process, with two connected tools for before and after your event.

Vision Board: Teams set their own custom questions about the event, like theme, mascot, or activities, and each member answers with just a word or two. As the team types, a live word cloud builds in real time, then those words automatically convert into a shared photo cloud, turning scattered ideas into one unified visual, built together.
Memory Board: After the event, teams can upload their event photos, and Memento automatically turns them into a beautiful storybook, preserving the memory of the event in the same visual, collaborative style.

How we built it

Backend:

Hugging Face's semantic model runs in-browser to compare response meanings and group similar answers (e.g. kayak + paddleboard + canoe) into one visual concept
A Python script controls tile size based on frequency, and more similar responses to a concept means a larger, more central tile in the photo cloud
Reve AI API generates custom images from each team response, and keeps a consistent illustrated style across the photo cloud

*Due to Reve limitations - Bing image search pulls cartoon-style thumbnails for each concept, using a styled query (e.g. "kayak cute 2D cartoon vector illustration isolated on pure white background") with a forced white background and fallback to a bundled local image if a fetch fails

Frontend:

Claude to help design and build the UI, focusing on layout and overall experience
Codex to help generate and refine frontend code, speeding up implementation

Challenges we ran into

Limitations of the Reve AI API
Time constraints, hence we have a few key future implementations

Accomplishments that we're proud of

Building an efficient semantic grouping system with Hugging Face that accurately clusters similar responses without needing a paid or heavyweight model
Getting the entire application working end to end, so you can actually plan a real event using the app from start to finish
Creating a useful, easy-to-use, and genuinely fun tool for hackathon organizers (like SummerHacks!), university clubs, and small teams planning events together

What we learned

How to work with AI image generation APIs like Reve, including their capabilities and limitations
How to use semantic similarity models to group related ideas together in a meaningful way
How to integrate multiple tools and APIs into one cohesive full-stack app under tight time constraints
How to turn creative, freeform team input into something structured and visual, without losing the spontaneity of everyone's ideas

What's next for Momento

Smarter AI-powered image generation to improve accuracy and visual quality for the vision board
Evolving memory boards that support more formats beyond static images, like GIFs and videos, to better capture the full experience of an event
More personal expression options, allowing users to customize and personalize their boards to reflect their own style and voice