Text to Speech (TTS) on Rockbox

Text to Speech (TTS) on Rockbox

This project is a part of Google Summer of Code. It is initiated by MarcoenHirschberg and taken on by OzgurOzturk.

We considered porting one of two open source TTS engines to Rockbox. One is eSpeak, a formant-based extremely light TTS engine producing less-natural/more-robotic sounding speech. Other is Flite, a port of Festival for mobile and embedded systems. It has a larger footprint, around 2.5 MB, since it uses prerecorded diphones to synthesize the speech from text, but the output sounds much more human-like.

When it works, you will be able to load simple text files of hundreds of books (for example from Project Gutenberg) to only a few megabytes of your player and listen to them.

Planned for this Summer

1. Successful port of Flite as a plug-in with the ability to synthesize simple text files
2. Including Flite to the core
   Making option available to make Flite read file and directory names and ID3 tags

Suggestions for Later (Please add your suggestions)

1. Adding filters for PDF HTML and other formats so that Flite can read them (maybe in collaboration with proposed DocumentViewer project)

The Problems

The voice files are included into the program as data structures, this requires large memory when compiling Flite and hence will cause the same for Rockbox when Flite is incorporated into the core. (We might decide to change this and have seperate voice files. TBD)

Progress

I have recently started working on the plug-in. Have successfully compiled Flite on my new set-up and my first (HelloWorld) plug-in.