Ekho - AI Journaling

Inspiration

At the 2014 Best Actor Oscar Awards, famous actor Matthew McConaughey said, "So you see every day, every week, every month and every year of my life, my hero's always 10 years away". This was the spark that led to the creation of Ekho: What if you could converse with yourself 10 years in the future? In a world where mental health and identity seem to be faltering away as social media and other phenomena take over, it is easy to lose track of who you are. So how do we maintain our unique identity and continue self-growth? Traditional journaling can be tedious, therapy is expensive and intimidating, and chatbots seem very detached. However, Ekho offers the ability to talk to a contextually aware and sentient version of you to converse with, whether it be about your day, your recent problems, your future goals, or really anything you want.

What it does

Ekho is an AI-powered companion that is personalized for you. It can become an older version of you, allowing for much more relaxed conversations and transparency. It has many modalities of functionalities:

chats with you to unpack your day
provides mood-aware responses to support your mental health
stores information to enable context-aware conversations (ex. reminders, goal setting, or chill chats)
data analysis for mood tracking and pattern recognition (You can see how you think!)

How we built it

Backend

Python
FastAPI and REST
Gemini for script generation and responses
Google Veo 3.1 for video generation of older self
Google Cloud Storage for storing reference photos and generated assets (videos and audios)
Google ADK for sentiment and mood analysis
Snowflake for long-term data analysis and user sentiment tracking
ElevenLabs for realistic audio generation (free tier, with pro tier can mimick user tone and make it more mature)
MongoDB for storing recent chats and avatar images

Frontend

React, TypeScript, Vite

Challenges we ran into

API integration

API integration always took a long time to implement, especially Veo. It took time to understand the documentation, and we often messed up the required inputs of the APIs. It took hours to make sure some of the integrations were working properly and not throwing errors.

Time Complexity

Loading audio and video files and generating them took long periods of time, which is quite inconvenient in the real world. However, it isn't really possible to get past this barrier.

Accomplishments that we're proud of

-We were able to take this very creative idea and unique experience into a real-world working idea. We were able to materialize our future older selves. -We were able to integrate a large tech stack to work harmoniously together to produce actual results that will better the world around us

We developed something that we, as a team, all personally connected with. We want to be able to talk with our future selves and just discuss our lives, goals, and ideas. Very proud of how we were able to turn this into reality.

What we learned

Small mistakes and misconceptions take lots of time. Oftentimes, it is difficult to pinpoint the errors and what's wrong with something. For example, it took us hours to figure out how to successfully generate videos with Veo due to complications with documentation and functions.
Finding a pain point that affects us is important. Not only do we know that there definitely is an existing user base, but we are also motivated to tackle this challenge.
We learned a lot about the various tools that we integrated, such as Veo, Google Cloud, MongoDB, SnowFlake, etc.

What's next for Ekho - AI Journaling

-We want to explore how to make the video and audio generation process more efficient. Maybe by splitting up videos or reusing past videos. -Incorporate research-backed analysis on sentiment trends and help users understand their own developments more. -Expand to the mobile app -Add additional ways for the user to interact with the app. Maybe it can turn into this companion thats always with you throughout your day and automatically helping you document.