For my TTS interaction, I have made a Node.js project that will give a user the weather when prompted. Read more below about it.
I originally wanted to find an API that would be able to take images, and output a description of the image. I thought that it would be something practical for visually impaired people, however I do know that similar things (like reverse image search and accessibility apps) already exist. I just wonder how it would work the constraints given to us for this assignment. I attempted to do something with the knn image classification in deeplearn.js, but I wasn't able to find enough documentation to figure it out.
Once I accepted that I wouldn't be able to accomplish that with my current technical skills and after loosing several hours to it, I decided to use the OpenWeather Map API. I thought that using a voice to convey data that changes is far more interesting than hard-coding a limited number of TTS responses. Going along with the idea of a voice interface being conversational, it makes a lot more sense that we would like to have responses that vary from time to time. While the weather isn't the most thrilling topic, I think it provides enough variability and functionality to keep things somewhat interesting.