ZDSR screen reader: questions

The Zhengdu screenreader, ZDSR, is a Chinese-based screenreader, which is likely why you have not found a whole lot of info on it. I found a little something on it, but this webpage is a mix of Chinese and English, so bear with it. However, it does give a little bit of info on the Application.
My humble advice to you would be, stick with NVDA.

https://www.qcsalon.net/ru/forum6/topic99466

I found this link to an earlier versino of the ZDRS manual. I think it is from 2021, but it should still give the basics. However, based on their payment model and terminology, I think I'd do as Brian suggests and stick with NVDA. After all, it is not until the commercial version that you get access to, as the manual says, "navigation keys". Also, for the higher paying plans, you have to have a dongle plugged into your computer at all times while using it. Maybe that has changed since this version came out, but that's what it says. Anyway, without further ado:
https://github.com/ZDSRInternational/zdsr-docs/blob/master/Documentation/zd.md

I have a cold so can't say much, I've tried it, it's very fast, it works on chrome browsers, but I still prefer NVDA.

Here's the link to download it: https://www.icloud.com/iclouddrive/025XMlx0nsQfTN0QmEOBsuGJw#zdsr_setup_1.7.0

Isn't this sr that is only downloadable from weird cloud provider like this very icloud link? Not because it's chinese but I can't really trust this sort of thing...

Let us know how it goes, should you decide to test drive this beast. :)

The safest would be to install it on a virtual machine (VM) if you have one.
I'd advise closing, and not visiting any critical sites like your bank, while you test drive. Until you are certain it is legitimate.

That one is interesting. Does anyone have any idea in what all areas it's better than or out-performs NVDA/JAWS? I understand it's very! responsive from my quite limited research...

I've tried it and there are no viruses at all.

If you want to uninstall it; it's as simple as going to remove programs, pressing z, going to your applocations menu, pressing enter on uninstall, pressing y twice, and you're done.

There is one more thing that is installed, it has no name for us English users but again, just go to remove programs, go to the bottom of the folder, applocations, uninstall, yes, and you're good to go.

I just won't use it because I'm not a fan of the voice but the speed is awesome!

Telegram Group:

https://t.me/zdsrgroup/

web sites (translation required):

https://www.zdsr.com/

https://www.zd.hk/

If there is an official English web site these days, I ain't finding one.

A screen reader needs to access all of the same accessibility API's that, say, a banking trojan does, in order to give the right information to the user, so hiding a peace of spyware in a screen reader would be a pretty neat idea.
I would never install a screen reader from an unknown or untrusted source, especially something that could potentially have ties to the CCP. Just because it's not flagged by anti-virus doesn't mean it can't be leveraged for bad, and just because you officially uninstall it doesn't mean it doesn't leave something behind. Be careful, and if you must, test in a VM.

Okay okayy we're getting in danger zone here :) :) , I'd just say that I never said can't trust this because is Chinese or whatever, just that if they don't have an official way to download it properly seems odd to me. I don't care who made it. Commentary is something I'm very curious to test one day for example and it's from the same region.

Perhaps because English voices can't speak Chinese? Anyway, it's truly more responsive than NVDA if you use SkyVoice. The voice has a brokenish accent and most features require you to upgrade to a paid version, but the screen reader does work, possibly as well as doing other harmful stuff unknown to me. I doubt that though, as I have been a member of that Telegram group for a fairly long time, and I even translated some earlier version into Turkish, but I eventually had to uninstall it due to some weird problems not caused by ZDSR alone but due to a conflict with other software installed earlier. The guy responsible for a bunch of tasks including testing ZDSR and promoting it in the West, working on English localization and updating the English manual, wouldn't tell me the exact cause or how to fix the issue. This rendered the license he gave me in return for the translation totally useless, and all I was left with was one more thing to brag about and add to my CV.

Hello Enes,

Now it has an English name also, and that refers to its functionality. It is a cloud-based app mainly interesting for Chinese users. It has a download area, some voice to text capability, news platform, some form of an intelligent voice assistant etc. One of my three voice schemes in ZDSR is a dual voice, so I know the Chinese name of it is pronounced something like the English words "the shone".
It is totally harmless, and if one doesn't want it, it can be uninstalled in the standard way without problems.
I have the latest ZDSR numbered 2025.1 with an online license. I like it very very much. I haven't experienced anything "bad" with it. As nothing in the world is perfect as we all know, it has bugs too, but it is extremely pleasing to use for me. E.g. there are some apps which are totally inaccessible for NVDA, but to ZDSR they look like totally ordinary software. And there are also dozens of features which I especially like and appreciate.
Last Wednesday I sent that guy you refer to (his well-known nick is Nuno) a large package of bug reports and feature requests I digested from the one month of very extensive use since I have my license. Mind you in that month there were two updates with some significant bugfixes, so the developers are very active on ZDSR, so I didn't need to report those. The current 2025.1 is considered public beta, that is probably why the download section on www.zd.hk and the ZDSR help system is not updated yet. But the detailed release notes always contain the new features, of course it needs translation.

For example there is a piece of software called "Detect it easy" (DIE). It is a file type checker and file content analyzer I use a lot. That's a QT UI type-of-thingie, that is very usable with ZDSR in both normal keyboard/tab navigation and in ZDSR-special navigation (one of my favourite ZDSR features). In general it seems that ZDSR handles the quite popular, but very NVDA-unfriendly QT UI apps very well, which is in itself a big plus, because programmers tend to love QT UI, so it is everywhere now.
I don't have any English manual as such. What I use is the built-in help system and the release notes file (.txt format) for the new features. Both are in Chinese, but ZDSR has for example an automatic translation mode. I just press ZDSR+F7 combo (ZDSR key is the same as NVDA key) and from that onwards I hear everything in English (translation engine is configurable in Settings, my default is Google with autodetected language to English). It is so quick that I more than once was shocked why some Hungarian websites (my native tongue) suddenly switched to English. Bingo, I left the automatic translation on. ZDSR+f7 again to turn it off of course. But there are other very handy translation combos, like ZDSR+b to translate the last heard speech. I use that many many times too.
I note that the built-in help is not up-to-date unfortunately, so there are some features I learnt through khmmm. interesting ways. Like ZDSR+h for speech history: you get a window of the last many (default 200) piecdes of spoken text. Or numpad 7 is continous reading in ZDSR-special navigation mode. These are not documented yet, though they are not entirely new. I am quite certain that all documentation will be updated once the current iteration of ZDSR is considered not beta any more. There was a big big update last November, and since then it is considered public beta.
And also since last November, there is no need for a dongle any more. You can buy one if you wanted, but I don't know why because the online license gives you the full feature set. And there are no more youth etc. version. There is one package (linked above by Brad, and can be updated to the latest 2025.1) and the features available depend only on the authorisation. Maybethe dongle is retained for those on a very poor Internet connection. After all China has very distant rural areas.

@Laslo could you also say something about what interesting features that ZDSR has that is not present in NVDA/JAWS?

But you have to pay for it.

I think if peple are interested, they should wait for a couple of months/a year, hopefully by then an English version of the manual should be out.

What really stands out to me are the following: 1. The speed. This thing is fast, I'm able to zip through my computer with no issues.

2. When I'm on youtube on microsoft edge; if I press alt right or left arrow to go through each chapter, it doesn't repeat it twice, for example chapter one chapter one, it just says chapter one. (I'll be reporting this to NVDA to see if they can do anything about it.)

3. The music. Honestly that's a really nice calming touch.

Whilst I don't think i'd buy this screen reader personally, I really like what they're doing.

So here are some of the ZDSR features that make ZDSR usage really joyful to me:
- ZDSR-special navigation: this is a bit like NVDA object navigation, but way simpler. Controls/elements can be traversed on only two levels: main elements and natural children (like list items of a list). No 13 levels of parent-child object relation trees as I often saw with NVDA. There are some further helpers for nasty and poorly designed software, but they are also very easy. Like there are two shortcuts to discover text hidden in controls which NVDA would simply report as unknown. And this navigation does not tend to go out of the app window as NVDA often does for not-well behaved apps.
- superb mouse handling: under-the-mouse reading is way way more reliable than NVDA both with simulated mouse or a physical one. If you use that ZDSR-special navigation and have found something interesting, you can click that right away, no need to route mouse pointer to it first.
In 2023 this superb mouse handling was a lifesaver for me on a very non-standard website where there was a strange multiselection list, which was totally hopeless with any other screen readers. And web developers didn't seem to understand at all, what the problem was with it :( That's when I began to consider moving to ZDSR full-time for the first time. And that was with the free version, so that is also quite capable for otherwise inaccessible situations.
In OCR mode you can click accurately to the level of characters. Also very handy for poorly designed software.
- speech review command set: too often we just have to use various cursors and navigation modes in e.g. NVDA just to spell something out we have just heard. But not in ZDSR. The piece of text we have just heard can always be browsed with shortcuts by characters, words, lines or sentences. The same can be done for clipboard contents.
- text can be selected by marking the start and end point in every text field (including web pages) by two special key combos.
- by default typed characters are echoed in password fields. This is configurable. But in a solitary home environment it is quite convenient, because otherwise I have always typed out my passwords in e.g. Total commander command line, and then copied them to where they belong. However I encountered some web password fields that didn't allow copying into them. ZDSR approach lifts many tedious password operations.
- image descriptions: it has every way you can imagine for doing this. You can submit an image with a specific question, or get a general description and ask follow-up questions. You can attach multiple images too. You can have the "operation object" described with one combo: Alt+numpad /. So e.g. you arrow down on a webpage and hear "graphic" with no alt text. Just press that combo. OCR works that way too, in fact image descriptions are considered a special mode of OCR.
- super simple chat AI dialog with Vivo language model: that does the image descriptions too. You have an input field, just type out your question and hit alt+s and then the answer is read out to you, and you can review it by shift+tabbing to the response field. That couldn't be simpler. And it is there all the time and understands many languages, not just Chinese.
- translation of course: I wrote about that above. It is just available in the paid version for a reason - they pay for the API keys of the translation engines (there are approx. ten of them to choose from). That way it works better than e.g. the translate addon of NVDA, which utilises the free Google API. More text can be translated and results are quicker, as of course translation engines prioritise paying customers.
- monitoring changes in almost any object or on a portion of the screen dinamically
- marking important elements with road signs to find them with fewer combos
- assigning a label to any control or element with just a single combo

Wow. This seems very interesting.

There is even much more for sure, but one particularly came to my mind and that is not to miss. ZDSR does not switch between interactive and browse mode "on its own" on a website. Despite turning off all relevant settings, NVDA does that "autonomously" more and more. I humorously call that behaviour "NVDA is blinking its eyes rapidly" - because the mode switching sounds coming in rapid successions allude to that to me. But that piece of humour is just to ward off the frustration this caused in NVDA and it has been increasing rapidly as web pages and especially web apps have become more and more complex. Take Google Drive as an example. If you have tried that one with NVDA, then you know what I am talking about. Now there's no "eye blinking" with ZDSR. And there is even a way to lock interactive mode if that is needed for a web app. Very handy with Google Drive for example.
And I mention a very trivial omission of a feature in ZDSR in order not to just be on the positive side :) If you press Pgup / Pgdn in a web browser that does nothing! There is no such movement on a webpage implemented in ZDSR yet! It is so strange! Obviously no ZDSR user must have really complained about that, because otherwise that would have been implemented extremely quickly as much more elaborate features / fixes had been. I have been watching the progress of ZDSR since 2015. Yes, there is alphabetical navigation (almost identical to NVDA), Zhengdu search (similar to NVDA virtual search with some twists) and a command to jump a block of links (default 10), but there are still many scenarios where none of these are appropriate. This is one ofmy feature requests in the Chinese lunar new year "package" I sent to Nuno. Oddly enough Orca under Linux behaves the same: the last time I checked (not recently I admit) it also didn't react to pgup / pgdn on webpages. But it is much less of a frustration with ZDSR than with Orca, as ZDSR is extremely quick even on a hyper big website during arrowing, while especially Orca (and also NVDA to a bit lesser degree) not so much. This is no surprise: ZDSR is written in C#, while Orca and NVDA are Python products. And the latter is a well-known example of an inherently "slow" language.
But again something very positive for closing this post: Facebook is way more agreeable with ZDSR than with NVDA! I used to humorously call Facebook "experience" with NVDA as "CIA torture degree 4". Now with ZDSR that is lowered to "torture degree 2". That is still not for the faint of heart, but it is still the best FB "experience" I have since FB mobile workaround pages are a thing of the past. The scrolling behaviour of FB seems to kill ZDSR the least. This is even more remarkable as the Great Firewall of China protects those citizens behind it from such experiences.

You can always disable automatic brows mode that's the first thing I do. Those who use a mac are not allowed to say that nvda object navigation is combersome as VO is 10x more, in my opinion. You may have enabled the layout table by accident which indeed increase the nested objects a lot. The first time I did that I was like what the heck is this. I get the point of this feature but aside very very specific usecases you wanna disable that.

When ZDSR developers changed the licensing system in last November, it brought about one further pleasant side effect. If you have an online license, you can freely transfer it to another computer. There are no more transfer fees and transfer procedures etc. If you for example changed computers, all you have to do is log in with your ZDSR id and password, and ZDSR will know about your license on the new computer. Only thing you must pay attention to is that you shouldn't use two computers actively at the same time with the same ZDSR account. And there is no more restrictions on using ZDSR on a virtual machine. Youth version used to have that, but it was "silently" lifted long before the new one-package format. I say silently because I happen to know it from therelease notes.
to BlindGuy: I didn't have either any automatic browsing setting, or layout tables on in NVDA. I disabled the former in the first days of my NVDA usage (way back in 2010, sigh) and I also discovered the drawbacks of the layout table feature soon, so I disabled that some months later. The "eye blinking" and the very complex object trees were beginning to get more and more frequent much later, around 2018 on onwards. This is related to more and more complex apps and especially web apps, more and more wrong coding practices, and more and more breakage of accessibility APIs.
As I see it from developer communication (on NVDA-side) and from other factors, NVDA developers believe in accessibility APIs more firmly and are (especially in recent times) more reluctant to hackishly work around scenarios where developers did not follow accessibility APIs. On the other hand, ZDSR developers do not expect developers to be that well-behaved, and they hack around problematic cases much more freely. That is mostly why ZDSR is usually much more resilient in edge cases, which are - let's be frank - everywhere. Though there are counter-examples also, albeit very rarely. There is an extremely poorly designed banking website I have to use regularly, which breaks every standard in every way, and by some miracle with NVDA it can be used almost entirely (albeit with significant tricks), while with ZDSR functionality is more restricted. That is the only scenario right now when I switch back to NVDA for a swifter workflow.
Me personally am for the concept of ZDSR regarding accessibility APIs, because I find that more pragmatic and realistic, and the concept of NVDA a bit idealistic.

Firstly, let me say that I am not down playing ZDSR in any way, shape, or form. I would, however, like to correct a couple of misinformed comments here.
1. Disabling auto focus on NVDA is the only way to go. I am not disclaiming the "blinking eye" issue that you all have experienced, I am simply saying that I do not experience this.
2. NVDA does indeed have hockeys for navigating forward and backwards by characters, words, and lines. Just an FYI there. :-)
3. Add-ons exist to sure up any weak points of the base code of NVDA. There are improvements for object navigation, mouse navigation and interaction, and translations. As well as a plethora of OCR (both AI driven, and non-AI driven) add on software as well.

Again, I am not in any way trying to downplay ZDSR. Just simply pointing out that NVDA is not anywhere near as bad as you are making it sound.

Thank you for reading.

Perhaps you had an older version of NVDA but todays should work with no issues what so ever.

The keys for letters and words and all that are the same in NVDA as they are in JAWS, control, left and right moves you word by word, shift, selects letter by letter and the left and right arrows moves you letter by letter.

Up and Down arrows navigate by lines. There is also a little add-on to allow Alt + up arrow and down arrow to read by sentence. 😜

What would be the cost of a fully functional lisence of ZDSR? I'm not asking about a comersial version, just for single use... And how responsive are the devs to custom issues? like a specific user facing an issue with a propriatory piece of software?

I do experience this annoying issue where I'm just kicked out of a text field on a web page. Do note that I am using NVDA 2024.4.1, but I highly doubt updating to 2024.4.2 will fix that. I'm just waiting for whatever big update awaits us in 2025. Anyway, I would like to know what features of ZDSR are available in the free version. I helped with Turkish localization but couldn't use the license I got and was not told how to fix the issue that prevented me from being able to use ZDSR. It wouldn't even launch properly at all so I'm not gonna pay a single penny unless the paid version significantly outperforms NVDA. So I would like to know what functionality the free version offers. Your posts make me conclude that the paid version is quite robust indeed, but the free version used to have some severe limitations back when I tried to install and use it. Is that still the case? What functions previously exclusive to the paid version have been introduced in the free version as well? Another question is, how is ZDSR so responsive with SkyVoice but not any other TTS engine? My computer unfortunately came preinstalled with a single-language version of Windows, which is Turkish. So I doubt I would actually set the display language to English solely in order to be able to use ZDSR more conveniently, but this is irrelevant now, as I won't be able to do so even if I wish to. I am quite used to having Eloquence read out stuff in Turkish, but this may not be the case for SkyVoice.

Glad to be of help. Also, you may want to disable the auto focus by carot movement, this is where you just use up and down arrow keys without using any other NVDA commands while navigating a webpage. That way you don't get stuck in edit fields this way as well.
Just a pro tip for ya. 😉

I reply to the several latest posts here in one go.
I also have "automatic focus mode for focus changes" turned off for at least 13 years now, like all automatic focus/browse mode switching options, because I experienced their detrimental effects in the first days of my NVDA usage. No, I never ever had focus problems with the Google search field. But I did have serious "eye blinking" in for example the share/create link dialog of Google Drive recently, up to the point I got stuck, and I had to work around that in various ways. I emphasize "recently", because say one year ago it didn't do that at all. Nowadays things important for a sccreen reader can be mucked up totally unintentionally in a matter of seconds anywhere and the fixing date is uncertain to say the best. I just state that for example with ZDSR I can do the share / create link dialog in Google Drive in seconds without issues.
I've been using NVDA for 15 years. I liked it for the first minute and I still do and it's really a great screen reader and also totally free, a very important point for many. Folks, let's be clear: humans differ significantly. Our computer usage patterns, our needs, our life situation, our financial situation, and totally subjective and irrational personal preferences - everything.
The current license fee for ZDSR is 26 U.S. dollars per year for an online license. It activates the full feature set, which was exclusive to the dongle version earlier. Now there are no "in-between" feature set versions, just the free one and the full featured one, no matter what licensing scheme you choose. In Hungary, which is a lower middle income country on a world-scale, this is what I pay in a very modestly priced restaurant for a simple lunch, or for two or three days ofhome-delivered regular meals. Taking all circumstances into consideration,ZDSR is highly worth its price for me. And I doubly emphasize "for me", that is as I said highly subjective.
Developers' responsiveness to custom requests: from the release notes it's quite clear that they actively listen to user input, however user input comes from Chinese users virtually exclusively. Let's be frank: I would be highly surprised if there were significantly more active ZDSR users in the "western world" than how many fingers and toes one person has altogether. I read regularly that they made adjustments topieces of software we never hear about in the West, like "Jimen trader" or "easy programming language".
A good test will be what happens to my large bug report / feature request bundle. My feature requests are mostly trivial to implement (require few lines of code) and not software-specific. It will be very interesting to see how this proceeds.
Free version limitations: there are still many, many of them, and a lot of them is totally arbitrary. The document on limitations is also out-dated, but from experimentation I know there are at least two former limitations that have been lifted recently. There are now alphabetic navigation support for Chromium-based browsers in the free version, and also laptop layout. There may be others though. But most limitations are in place: no Firefox support, no translation, no OCR, no Java support, no advanced "professional software" special support (only basic one): Word, Excel, Powerpoint, Reaper, Sonar etc. The latter means they behave just like regular programs, and the plethora of supporting special ZDSR hotkeys for them are not available in the free version.
Speech synth responsiveness: I use Espeak Hungarian through SAPI5 most of the time, and I can definitely say it is very responsive now, absolutely no problems anywhere. That has come a long way, because it used to have significant problems there till around 2022/23. That's why I waited so much time for ZDSR to be my daily driver. Now I can confirm it is superb in the free version too.

Well, the auto-focus thing comes in quite handy in numerous situations, so I'd rather have it fixed than have to disable it myself. Perhaps NVDA should wait for the whole page to load even if eventually does jump to an edit field, which is also quite convenient by the way. Visiting a search engine page and being able to type in your query in mere seconds is quite practical indeed. As for the limitations that the free version of ZDSR has, I would actually use it from time to time if I had a portable version,, but now I don't know what weird issue I will face so I just refrain from installing it in the first place. The last time I had such issues, I was not even told why they occurred and how I could fix them, which caused my well-deserved license to go to waste. This is why I don't really want to pay any fees whatsoever, not even a scent. I won't even be able to test how the other TTS engines stand against SkyVoice now in 2025.

Hi,

You can disable NVDA from automatically jumping to an edit field, or focusable item, on page load. Just FYI. 🙂

I conclude this will actually prevent NVDA from switching back and forth between Focus and Browse modes while a page is loading and/or once loading is completed?

As I understand it, you will still automatically enter an edit field as you pass over it, with auto focus enabled. The setting I am referring to, just makes it so that NVDA does not automatically jump to a focusable item as soon as a page or app or whatever first loads. I hope that makes sense. I can't remember exactly which subheading the option is under, but if you Poke around in your NVDA settings, you will find it.

Thanks...
Anyway, back to topic. So I doubt we can use SkyVoice in other languages apart from English and Chinese, and that we can obtain the same level of responsiveness with another TTS engine. I am wondering though, whether SkyVoice has a SAPI5 version or can be installed as an NVDA add-on.

I think it's the screen reader itself.

NVDA can be fast when you install it for the first time but it does seam to slow down, or that could just be me, the older your computer gets But zdsr is fast and very responsive.

I would ask the NVDA devs to take a look but I don't think they'd be willing to change whatever it is so they both are as fast as one another.

ZDSR is so fast that when I press enter on a youtube video in edge, it loads in less than a second, I'd not be surprised if NVDA does that too but it doesn't feel that way to me.

I think all screen reading software gets sluggish over time. ZDSR will too, eventually. Not putting it down, this is just the nature of current software archetecture. Everytime you do a thing with your screen reader, a cache is stored somewhere. Something is indexed, something is saved for that very next time you need that thing done.
Over time these small bits of data begin to drag a program down.

Though in your specific case, Brad, it might be a good idea to back up your profile, and reset NVDA, then reload your prof.

Just a thought ... 🤷

It's hardly sluggish these days but thanks for the suggestion.

That'll be $16.95 for services rendered.
Greetings from your friendly neighborhood AppleVis Tech-support. 🤣