Speeq

Inspiration

During the opening ceremony, we noticed that many of the speakers went overtime. We wanted to help people keep track of time while public speaking, so we developed a tool that tells speakers how much time they are estimated to take and how much progress they have made through their pre-uploaded transcript.

What it does

Speeq listens in real-time to speakers and calculates the rate at which they are speaking to help keep them on track to meet their target time.

How we built it

We used python to repeatedly record short audio files of the speaker and parse them using the Google Cloud speech to text API. We then used this information to calculate a predicted duration and constructed a basic GUI using TKinter to provide real-time feedback to the speaker.

Challenges we ran into

The biggest challenge by far was interfacing with the Google Cloud API. We originally planned on using Node.JS and building out a web app, but switched to python at the last minute after finding this to be infeasible due to our teams lack of experience with Node.JS.

Accomplishments that we're proud of

We developed a strong minimum viable product that has all the key functionalities we set out to implement. We are all very proud that we stuck it out and got it done, because we sat still at the bottom of a steep learning curve for a long time. Without this determination, we would not have a product to present.

What we learned

We learned the basics of Node.JS and how to use the Google Cloud APIs to perform text to speech analysis in python. We also learned the basis of various skills from workshops, including webex, react, google cloud, and spatial computing.

What's next for Speeq

We want to continue to work on this product by building out a better front end and improving the latency of speech recognition, and incorporating better NLP and machine learning applications. Our mission is to make a useful product that can ensure CalHacks 7.0 starts on time :)