Summary
| Mentor(s): | Dr. Mike Murphy |
|---|---|
| Project Type: | Software development |
| Technologies: | Python, HTML, AI Text-to-Speech (likely Piper), Git |
| Undergraduate Credit: | CSCI 399, CSCI 490 |
| Graduate Credit: | IST 659, IST 799 (Thesis) |
| Locality: | Main campus and/or remote |
Purpose
TealSite produces HTML documents as its primary type of output. HTML is processed by a user agent to produce some kind of representation accessible to the consumer of the site. For sighted users, the normal user agent is the web browser, which renders the site in a format suitable for visual consumption. However, there are other categories of user agent, including those that read the text aloud for sight-impaired users. A disadvantage of a user agent that performs its own text-to-speech conversion is that the content is read in a generic voice. For instructional content in particular, this voice can make the presentation feel impersonal.
Modern AI-powered text-to-speech systems like Piper make it possible for a site author to produce a model of his or her own voice. With such a model, the text-to-speech engine could read the page content in the author's voice. However, it would be undesirable for most authors to publish these models directly, since they could be easily misused by bad actors to impersonate the author. Instead, TealSite could run the text-to-speech generation from the relative safety of the author's own computer.
Scope of Work
Automate the process of generating spoken text from the written text on a page, and integrate this automation as an optional feature of TealSite. To support major desktop and mobile browsers, the MP3 output format should be used. Add custom processing tags to the TealSite Python code to trigger audio generation and embed the audio for each page.
Basic text-to-speech support for TealSite would be suitable as an undergraduate project. More advanced text-to-speech work, including testing and developing a good process by which the author can generate a personal model, would be suitable for a masters thesis.