Program a tool helping Petra interacting in virtual meetings easier!
Image desciption: A smiling women with grey hair wearing glasses and purple blazer in front of a green natural background.
About Petra
Petra has multiple disabilities (speech, walking and visual). She uses assistive devices on a daily basis, such as smartphone and laptop with speech output. In her free time, she develops apps for Android, listens to audio books and also enjoys reading books. Petra works part-time for the foundation "Access for all" as an accessibility consultant. In this position she tests websites and apps for accessibility.
Challenge
For Petra, who has a speech handicap, it is challenging to communicate via audio channels in today`s ubiquitous media for both workplace and private settings. Especially in the workplace context, claiming a professional stance towards previously unacquainted peers or customers is difficult. The goal of this project is to develop a software for professional call management with a natural Text-to-Speech engine that easily integrates into today`s professional collaboration solutions like MS Teams etc.
Solution
We have split the challenge of Petra into four parts. The first part was to develop a solution to integrate the audio channels correctly. The sound of a text-to-speech engine and the sound from Petra's microphone should be the inputs into Teams. The text-to-speech sound should also be the output in Petra's headset together with the sound of the person speaking with her on Teams. Luckily there was existing open-source software, VoiceMeeter and VirtualCable, which helped to achieve the required audio routing. The second part was the integration of a text-to-speech engine, for which we implemented the possibility to chose between the Microsoft speech API and Google Cloud TTS, both with English and German voices. In the third part we developed a graphical user interface (GUI) which allows Petra to select text blocks, ordered in categories, which are then read by the text-to-speech engine. In addition, Petra can directly write text in the GUI and adapt the text blocks and categories to her needs. The forth part introduced autocompletion for the GUI. When typing in the freetext field, the prepared text blocks are dynamically filtered for quicker selection. In addition, the text blocks are ordered according to the frequency of their use.
On Sunday afternoon we managed to put all the parts together and test the solution with Petra on a Teams call - and it worked! After the Hackathon the team is still working on the challenge to provide documentation and an easy-to-install package for everyone in need. In addition a version of the software for MacOs is in development.