The Gear Grind

• 

  Window of naming section of the project

Inspiration

Growing up in such a diverse environment such as New York City, we often find ourselves surrounded by people of all types of backgrounds and ethnicities. We noticed a common struggle that we all had, which was mispronouncing or forgetting names. Since one of us had heard about ElevenLabs before, we decided to take advantage of it's features to build our product.

What it does

Users are able to interact with a character in two different ways. One is by having the user pronounce the displayed name on screen, the other is by having the user listen to the name and type it. The character would then react accordingly. There is also a point system give incentive to the user.

How we built it

We used ElevenLab API to integrate their Speech-Text and Text-Speech endpoints.. Speech-Text was accomplished by using the sounddevice packet to record the user's dialogue, which was then converted into text by ElevenLab. We would use the outputted text to compare it with the name and determine their similarity. Text-Speech was made by saving a generated audio file and playing it with pygame. The main language we used is Python.

Challenges we ran into

This was our first time integrating an API into a program, as well as our first time working in Python. Due to out lack of experience, a number of discrepancies popped up during our hack.

Accomplishments that we're proud of

We were able to successfully integrate ElevenLab into our project and create functional methods that align with our intended goals. A bonus was creating a GUI with python that acts dynamically with our code.

What we learned

We learned how to implement the Elevenlab AI. We also learned how to work in a Python environment and created a GUI. Additionally, one of our members was to able to persevere through 2 hours of tedious work and come to the fruition of downloading the language Python.

What's next for The Gear Grind

Currently, we only included typical American names due to the restraint in time and feasibility. In the future, we would love to include more a international set to better reflect the standards of today. We would also like to use a better trained or custom trained model, and possibly integrate Opik into our design to increase quality in our data. We are also thinking about using a different form of front-end since we didn't really know what front end to use, and we settled on one which didn't have much customization. We also want to be able to provide feedback for incorrect answers and be more supportive to our users

Built With

• elevenlabs
• python
• visual-studio