Hello
If you want to make your own voices for nvda here’s how to go about doing it. to use the voices you make you will need the nvda festival drivers and the nvda screen reader, preferably the latest main snapshot or display hooks snapshot from www.nvda-project.org
Josh wrote:
> Hi
>
> I have some questions about the festvox project.
> 1. is it possible to create festival voices under windows? if so, how? if not could I use knoppix-adriane in a virtual machine?
The voice build code does run successfully under cygwin under Windows,
though it’ll probably be much faster under Linux.
> 2. can voices be created for the flite sapi5 speech synthesizer in windows, if so how?
Clustergen (statistical parametric voices) are well supported for
conversion to flite voices for English. For other languages you have to
build the text support yourself in C.
> 3. Could I use a patch cord to record rare or older speech synthesizers and create festival voices based on these speech synthesizers? examples like apple alex, apple pipe organ and others.
You would be responsible for the legality of that. Technical this can
be done, I have build synthesizers from recordings of synthesizers
before. It mostly works, though there may be some surprises.
> 4. is there a sapi5 version of the cmu-hts voices?
Note the sapi5 support in flite almost certainly doesn’t work. It
hasn’t been used for years and only still exists in the distribution so
that it might help someone update it.
It depends what you mean by “hts” HTS is a different project
distributed by Nagoya Institute of Technology. CMU Flite does not
include any support for HTS developed voices directly. Clustergen is an
alternative method, but the best HTS voices are better (but may have
licensing restrictions). I know people have successfully got running
HTS voices with flite, but I don’t have access to any of that code.
> 5. where do I find the 1300 or so phrases the speaker or recorded synthesizer would have to read?
> I want to make all my voices for u.s. english.
http://www.festvox.org/cmuarctic
The instructions to build a voice are in a homework I set
http://www.speech.cs.cmu.edu/15-492/homework/hw1/index.html
Don’t expect it to work perfectly, there is still quite an art in
building a good voice.
Alan
Josh wrote:
> Hi
>
> Is building voices accessible with linux screen readers such as orca?
I don’t know what orca is. Often linux screen readers use either
festival or flite, but they maybe offer other support too.
> will it work in knoppix adriane, or vinux? and in windows will the tools work with NVDA or system access screen readers? Since I am blind the tools have to be accessible with screen readers. I’ll mostly make voices for american english and british english.
>
I believe flite as is doesn’t work with SAPI, and hence wont work with
any of the Windows screen readers as they usually go through SAPI or do
not offer any changing of the voice.
Under Linux there is no “SAPI” like standard, but often they are using
flite as the example voice.
Alan