~Wavelength~

Companionship is resonance. Find your frequency.

Comment

Inspiration

There's something more than shared interests or MBTI traits that lead to lasting friendships and companionship. From witty banter to tuned emotional sensitivity, best friends (or more ;) ) have an established, deep rapport that arises from the first click, as if by magic. We've also seen the other side of it all-- namely relationships that lasted less than a week because the two parties just couldn't find an aligned resonant frequency in their conversations.

We believe that this compatibility and comfort arise from our internal personas and values, which are reflected through signals from our conversations, interactions, demeanor.

We want to create a platform for such friendships to be founded, where the barrier of improbable timing is eliminated, so that the door to longest-lasting relationships can be unlocked. Using artificial intelligence to analyze a short conversation, we aim to analyze and utilize the traits that are overlooked in friendship-forming applications: the way we speak, the way we don't speak, and the traits we may not know we have. After all, communication is essential to any relationship, and can make or break whether we enjoy someone else's company or not.

What it does

For everyone who is in need of a friend, or just wants to add some spice in their life by meeting someone new, ~Wavelength~ is a friendship matching app that unveils your true personality through a conversational environment to find others who will just click with you on a deeper level. Based off the YC startup "Dating Ring" (Winter 2014), ~Wavelength~ uses artificial intelligence to analyze how humans think and interact in order to determine their next best friend.

How we built it

We implemented our backend with Python FastAPI and Uvicorn. We used Google Generative AI's model Gemma 4 and ElevenLabs for text-to-speech conversions. For logins and user information storage, we used SQLite. We implemented our frontend with React, and built it with Vite. Getting started, we developed a basic foundation for the application together before splitting tasks based on our strengths and what we'd like to work on!

Challenges we ran into

One of the challenges we ran into was the latency when prompting Gemma 4 to respond to user's chat messages. The pauses in conversation made it unnatural, and we spent a lot of time to reduce these gaps as much as possible during the limited time we had. A second challenge was the ability of AI to determine the tone of the speaker during voice chat in order to accurately detect a user's unconscious mannerisms.

Accomplishments that we're proud of

We're very proud of being able to tweak the voice generated by ElevenLabs using Python libraries to make it sound as human as possible to inspire casual conversation. Additionally, we are proud of brainstorming, planning, developing, debugging, and marketing our idea from scratch in the time limit we were given!

What we learned

We learned that it is extremely important to research feasible features and drawbacks before jumping into the code. We realized that the AI softwares we incorporated into our code did not always work as expected, and thus we will definitely explore all possibilities and obstacles of each feature before implementation.

What's next for ~Wavelength~

We plan to further develop our site with more features and expand to more users! For features, we hope to be able to integrate video calls with matched friends. We hope to gather feedback from potential users as well, in order to understand what they would like to see in an app like ours, and what excites them about using it!

** Note: None of the descriptions above were AI generated (really!) **

Built With

Share this project:

Updates