[DECtalk] question about Sonata voices for NVDA screen reader

[Blake Roberts]

Hi everyone,

Although this thread is not about DECTalk, the focus of my message is a 
speech synthesis project. Starting in 2024, I have been using an add-on 
for the nVDA screen reader called Sonata. Sonata was previously known as 
Piper until a project name change. The add-on uses voices powered by AI 
to provide text-to-speech for the NVDA screen reader in real-time. Here 
is a link to the Sonata NVDA add-on page at GitHub:

Releases · mush42/sonata-nvda 
<https://github.com/mush42/sonata-nvda/releases>

I genuinely appreciate that earlier this year, the NVDA add-on developer 
created a version of the add-on which is compatible with 2025 releases 
of NVDA. I am someone who enthusiastically uses AI tools, such as 
Sonata,  when I have a purpose for using them. Unfortunately, I 
experience a consistent voice installation problem. I'm writing to 
ascertain if any members of the DECTalk listserv can offer a solution to 
the difficulty I encounter.


     Description of Problem: I'm running Windows 10 Home and NVDA 
version 2025.2. The most recent version of the publicly available Sonata 
NVDA add-on, 3.1.0,  is installed. The voices made available through the 
Sonata Voice Manager in NVDA menu, accessible through the "download" 
tab, consistently fail to install after download using the current 
Sonata release dated June 23, 2025. After the download of various voices 
in Sonata Voice Manager reaches 100%, I am told: "download failed". My 
Internet connection functions properly. Additionally, the "refresh 
voices" button also does not work.


In June of this year, I informed the Sonata NVDA add-on developer about 
these Sonata Voice Manager issues through email, and received no 
response. an updated version of Sonata NVDA add-on is not currently 
available. The previous version of the Sonata NVDA  add-on from 2024 
functioned without download issues. To my knowledge, the older version 
is unavailable. Even if I could download the 2024 release on GitHub,  it 
would probably be incompatible with 2025 releases of NVDA. NVDA 2025.1 
was a compatibility-breaking release, meaning incompatible add-ons 
cannot be installed.


Blake's Question: Is it possible to manually download Sonata-provided 
voices from a website, for installation using the "install from local 
file" item under the "installed voices" tab of Sonata Voice Manager? I 
know that installing Sonata voices as downloadable tar.gz files works 
because I installed Tamas's KeyNote Gold voice yesterday evening, which 
he created  using AI. I downloaded Keynote Gold speech synthesizer for 
the Sonata NVDA add-on here:

Piper voices, and other retro NVDA speech synthesizers 
<https://eurpod.com/piper-voices-of-retro-speech-synthesizers/>

The Keynote Gold voice file archive for manual installation into Sonata 
NVDA add-on is available  under the "Voices List" heading at the link I 
provided above. I appreciate all of Tamas's hard work to make this. 
Although I notice this emulated Keynote Gold voice sometimes cuts off 
the beginning of words, some voices offered by the Sonata NVDA add-on 
project do the same thing. I know from experience that some newer Sonata 
voices are less prone to start of word cut-off.


Final Thoughts: I am interested in also using voices provided by Sonata 
in the Sonata NVDA add-on. Being unable to download voices directly in 
the Sonata Voice Manager is, from my perspective, a problem. Any 
resolution suggestions which the DECTalk list members can provide would 
be helpful. Thank you for reading.


Blake
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://bluegrasspals.com/pipermail/dectalk/attachments/20251027/131b1609/attachment.htm>