Mbrolax is new text to speech engine that uses MBROLA speech synthesizer to produce speech.
(scroll down to download Mbrolax)
Mbrolax is completely standalone, meaning that it has its own text processing mechanisms and audio output systems.
It is also going to be cross-platform and open-source.
This is an add-on for NVDA screen reader that allows NVDA to use Mbrolax TTS engine directly.
This means that you're finally done with using MBROLA synthesizer in conjunction with eSpeak over SAPI5.
Also, there is no more blocking licence problem.
Mbrolax is very responsive and produces good quality, almost naturally sounding speech.
MBROLA synthesizer is used to synthesize speech.
Text is preparedin a given language to be synthesized with a given MBROLA diphone database.
In the end, this text is passed to MBROLA in PHO format and returned audio stream is reproduced
with Mbrolax's audio output method for current platform. In this case audio output relies on NVDA nvwave module.
Mbrolax supports contextual natural language processing with simple interpreting language in which you can specify preprocessing rules
which will be applied on input text.
Also, it is easily extendable.
Anyone can add new voice to Mbrolax using the MBROLA diphone database and Python programming language to explain to Mbrolax how to convert text to PHO string to be used with that database.
Documentation for doing this will be available on this site as soon as possible.
This add-on comes with following MBROLA voices:
* Croatian [cr1]
* British English [en1]
* US English [us1]
* US English [us2]
Note that US voices are temporary a little bit out of shape (wrong pronunciation, incomplete context rules apply, etc.).
In the future Mbrolax4NVDA will have an voice manager that will allow users to install and update newly created voices safely from this site.
Upcoming voices: Spanish and Italian..
* When a word is very big and uninterpretable like some gibberish longer than 40 letters
Mbrolax doesn't know what to do with it and sends its raw pronunciation to MBROLA synthesizer.
This may result with overflow of MBROLA internal buffers. In this case NVDA will stop speaking, and log will show this error.
To solve this, you must go away from the word that causes problem and restart NVDA using keyboard shortcut.
This will be solved in the next version.
* Special note is that Mbrolax is not tried on logout screen nor on any other secure screen.
It is strongly advised not to try to set it as a synthesizer for those screens.
If you are feeling adventurous, and simply have to try it, then first make sure that you have a sighted person beside you.
If everything blows to dust you will be able to solve the problem only when booting to safe mode and changing NVDA configuration manually.
* Sometimes inflection pitch changes can be a bit drastic.
* In English: some compound words and words that are not in dictionary aren't pronounced well.
* Sometimes there is a notable pause gap between words, especially when they are short.
Otherwise Mbrolax is very stable and usable.
Download and try Mbrolax4NVDA 1.0B 21.7 MB
Once again: this is only beta version, so do not expect miracles!
Copyright © 2015-2017 by Dalen Bernaca