I can't tell you how mutch I don't want this!

I know there will be some smart arse out there saying "Just install Linux" Pleas don't I have to use a screenreader called NVDA to read the screen to me as I am blind.

There is a screen reader in Linux but it just is not that good. If it was better then I would think about it. I have tried!

I am immensely sorry to hear your experience. What is lacking? I totally believe you that this is the case, I'm sorry.

Everything is lacking.

Wayland hasn't even stabilised their accessibility hooks, and in the name of privacy have undercut what accessibility tools can see.

X server has always had an awful accessibility story. The server can break and swap node handles as you're using them.

It's a real pain that accessibility features are always integrated into proprietary OSes first. Like the live captioning feature in Windows 11 (for the hearing impaired), it wouldn't be hard to implement it on Linux with Whisper, but it still hasn't been done.

You can also try Windows LTSC. A little bit more fiddly to set up than normal Windows, but, you get a break from normal Windows. You'll have no problem since you tried Linux as well.

Maybe you could try to figure out linux TUI/CLI stuff with a braille terminal? May not help with some websites.

NVDA looks like it is open source, it shouldn't be too hard to port.

> NVDA looks like it is open source, it shouldn't be too hard to port.

Yup. Just gotta invent a Win32-compatible Wayland first. This... Is sorta a "whole fucking owl" moment.

> As explained above NVDA relies heavily on Windows specific API's and cannot be converted to run under Unix based systems without a lot of work. Given how small NVDA development team is spending time on making NVDA work under Linux at a level at which Orca works currently would take years and mean much less development for the version for Windows. In short the more reasonable course of action is to spend time on improving Orca or other Linux screen readers rather than porting (which in practice would mean almost rewriting from scratch) NVDA to run under a non Windows system. [1]

Accessibility in Wayland is still in staging. [0] There is not the APIs you need, to port anything to using them.

X-Windows only supports Class 1 info over AAC. Class 2 was only ever semi-implemented, and is the more important class of information for the user. You basically need an Optacon, and too bad if you don't want tactile.

NVDA does work under Wine! But only with well-behaving programs running under Wine. It won't work for the rest of the system.

Gnome's Orca only works with Gnome-aware apps. It is supposed to work with Plasma for KDE things, but its a dice roll. It works with Firefox, Chrome, etc, because they go out of their way to make accessibility work better.

But Orca is about half as decent as JAWS or NVDA. Its a step ten years backwards.

Voxin (paid) used to work well, but seems to now be unmaintained. Certs expired, no updates for two years, etc.

[0] https://gitlab.freedesktop.org/mwcampbell/wayland-protocols/...

[1] https://github.com/nvaccess/nvda/issues/13196

Just don't opt in to this then? Nobody is forcing you, to go to the settings app, go to AI settings, go to experimental settings, and manually turn this on.

Yes, because MS is known for respecting the user preferences and not forcing anything even if you disable it explicitly.

You can try apple stuff, i don't know how good their screenreader is but I assume better than the linux one.

Nope. It ranges from same to worse.

VoiceOver is... Well, it has some AI layers that can sometimes rewrite the text it is reading. So... Think AI subtitles, but interacting with them.

JAWS and NVDA are basically Windows-only, because no one else has a decent accessibility story.

OpenWRT for the win!

You should post on the NVDA email list. https://nvda.groups.io/g/nvda Or the Screen reader list: https://winaccess.groups.io/g/winaccess FYI blind people do not like any lag when reading that’s is why so many still use eloquence and espeak.

Yeah blind people love eloquence

Sounds slow and like something from an anine

Speech speed is always a tunable parameter and not something intrinsic to the model.

The comparison to make is expressiveness and correct intonation for long sentences vs something like espeak. It actually sounds amazing for the size. The closest thing is probably KokoroTTS at 82M params and ~300MB.

I think he meant overacting typical for English dubs.

The voices sound artificial and a bit grating. The male voices especially are lacking, especially in depth: only the ultimate voice has any depth at all, while the others sound like teenagers who haven't finished puberty. None of the voices sound quite human, but they're all very annoying, and part of that is that they sound like they're acting.

I heard a little DVa from Overwatch.

The only real questions are which Chinese gacha game they ripped data from and whether they used Claude Code or Gemini CLI for Python code. I bet one can get a formant match from output this much overfit to whatever data. This isn't going to stay up for long.

I am blind and use NVDA with a sinth. How is this news? I don't get it! My sinth is called eloquence and is 4089KB

Does your Eloquence installation include multiple languages? The one I have is only 1876 KB for US English only. And classic DECtalk is even smaller; I have here a version that's only 638 KB (again, US English only).

2018?

I loved Skype back in the day. Me and my wife would use Skype kasts and chat to people from all over the world. It was good with my screen reader all so. After the buyout it all went down hill. O yeah and teams is a fucking pile!

Skype went downhill for me after it was acquired by Microsoft with many paper-cuts like account management and constant re-designs.

Trumpy is blaming it on DEI!