January 27 2019

Speechvas

Let your voices be commemorated. Get the undertones of your meetings with this artful new technology from Prototyp.

The idea

In a meeting there is often more information to take into account than what has been explicitly said. With Speechvas our vision is to capture the undertones using algorithmically generated abstract art. In real time, Speechvas draws a picture on the screen that represents the voices of the people in a room. By extracting sound parameters such as volume, pitch and amount of background noise, a user will be able to see a piece of art develop as the meeting progresses.

How did it go?

We managed to finish a prototype of a web application. At the frontend we added a GUI to control the coloring, thickness and curve of the lines, experimenting with different looks. At the backend we added a server using the SciPy scientific toolkit for data processing and the Sanic lightweight python server framework for communication. Once the server is started, the conversation is continuously recorded, sending processed data to the front-end over WebSockets.

If we had more time

If we had more time, we would play around more with the visual expression. Maybe we could add some additional features like sentiment analysis and being able to detect different speakers.