Haiku needs a screen reading utility

[richienyhus]

All modern operating systems now come bundled with a basic level of accessibility features, namely a screenreader. While these features are out of scope for Haiku R1, these features should be within the scope of Haiku R2.

Screen readers are made up of three parts: A backend Speech synthesis system An API for 1stparty & 3rdparty access to the Speech synthesis library An OS bundled screen reading utility for accessibility

For instance, ​Orca is the Gnome screen reader that communicates with the Accessibility Toolkit via the AT-SPI API. In Android you have Google Text-to-Speech, which is used by Google Talkback to provide speech feedback on what is being enacted, displayed or selected. In Windows, Microsoft Narrator uses the Microsoft Speech API.

BeOS did have 3rd party applications, TalkBox used the festival synthesis library and ​SpeakIt which used a database of spoken words. The author of TalkBox has offered to open source it in the past. However before creating or adapting an utility, we need a speech synthesis library for it to use.

Library	Language	Licence
​espeak-ng	C	GPL
​festival	C++	MIT
​EkonFain for CKJ	C++	GPL
​praat	C++	GPL
​CMU's flite	C++	MIT

Both 'Festival' and CMU's 'Festival Lite' use the ​FestVox dev tool. These two would be the best choice, with the lightweight flite as default and hevywight Festival swapped in if needed.

[pulkomandy]

mmu_man started investigating this and IIRC made a talk about it. Maybe he can document his results here.

[richienyhus]

There is also this speech synthesis engine written in C and under the MIT licence: ​https://github.com/syb0rg/Tritium

[mmu_man]

You can find ​the slides for my talk at RMLL2015. Another possibility than Orca is NVDA, but this one is currently Windows-only.

[mmu_man]

Just noticed ​this french article that lists new TTS engines.

[pulkomandy]

Moving to Unscheduled because there's nothing here that requires breaking binary compatibility as far as I can see.

[bagen55596]

Spam