Roku TV Voice Remote

As any good lazy slob knows, the worst part of my yearly GoT binge is having to reach for the remote to skip the intro. I tried to fix that issue by cutting out the button press.

This Python script allows you to control your Roku TV completely by voice through keywords. The program waits for the obvious commands "pause", "play", "left", etc. so you don't have struggle with the different UIs of different streaming services. Through language interpretation and your home wifi network, the program sends commands to the Roku API.

• Source Code

Built Using...

• Python

• Roku API

• Kaldi Speech Recognition

This project was a lot of fun to make. I had never worked with language models before this. Being from Wyoming, I generally think of my Western American English as having the newscaster's "no accent". After seeing what Kaldi recognized my words as, I no longer think that's true.

I learned the how to install, configure, and develop on top of large existing codebases such as Kaldi. After considerate time following README directions and command line output, I was able to integrate into the speech recognition software.

This also allowed me to refamiliarize with Python as I don't get to use the language regularly. It is very easy to use, but after dealing with other languages, the lack of parenthesis seems wrong.

What's Coming Next...

Continue to explore different voice models and find the one that allows for the easiest pickup for American accents. If made available to a wider audience, create a way the user can select their own speech model.
Set this script up to continuously run on a microcontroller with a microphone, always picking up as you watch tv. Then you would never have to even start and stop the program.
Currently, Kaldi listens in "chunks," so it is possible that the program stops listening and processes mid-word. Need to rework how the program grabs samples and interprets them.