Y Speak To Me

I caught the common cold before entering the hackathon. With all the socializing and lack of sleep, my voice began to disappear. I needed a way to communicate. My voice didn't work. So, I built a text-to-speech engine.

My system downloads Youtube videos and their captions. From the captions, it extracts time periods where a series of words are said. That information is fed through a Java program with Sphinx to align the words with the video. All this data is used to build small video clips of individual words. When a user inputs text in the website, a clip for each word is randomly selected. The words are stitched together, and then a video is born.

It also works as an API simply use http://yspeakto.me/build_video?text= replace with the message.

Share this project: