Speech Central Issues

Thanks for bringing the area for possible improvements.

First I do understand if someone just updated the app and felt like that sharing of files was dropped. However if you have read what I have explained that is not a proper wording to describe the situation, the proper wording is that it hasn't been successfully implemented yet despite the effort.

Second I have tried the bookmarks and it works well with VoiceOver on the Mac. I am not sure if you try to access them in the Annotations sidebar, but it works as expected from there.

Third the app supports a diverse source of content. As such more specialized apps that target just one or few similar file formats will always provide a better experience for that specific file format. I don't deny that there can be improvements, but the architecture of the app wasn't built for some specific file format and may work by a quite different logic than what that format is intended to by.

Finally I don't put any secret that 97% of my users on Apple's platforms aren't blind (on Android it is a bit less like 90%). I have received a lot of help, advice, encouragement and many other things from blind users and I try to give back in various ways - from making the app free on iOS and Android for them to enhancing the accessibility whenever possible. There are usually creative solutions to make the app work great for everyone. However if the decision has to be made on some specific topic on whether to make it work well for 97% of users or for 3%, the choice will be to accommodate to those 97% at least until some better solution that can work well for everyone is found.

Just a hunch, but maybe the book / document format? With some pdf books, i've noticed that the find, feature is... slow as hell. I mean in typing, but when I thy to use the function on say, epub books, it works without problems, so its something with the pdf book, for some reason. Maybe it’s the ;/ame with the bookmarks? Just a hunch.

Through these years I became fairly proficient for VoiceOver for iOS, but I can't say the same for the macOS.

As far as I can tell on the Mac you are allowed to go the level below the button, and if you do so you can't use the button anymore, so that could be the source of your problem. I'll try to disable this behavior to go below the button level, it should be possible.

Regarding the find, I can confirm that initial version wasn't properly optimized for large documents. Even then frequently it would work, but I was pointed to this by some users so certainly performance issues did happen.

This was resolved, but I can't say with 100% certainty if it was published in the current version for the Mac (on iOS it was). If it hasn't then it will be part of the next version.

Find has been changed completely to be more accessible.

Is there no way to read bookshare books with this app? If not, is this feature gone for good?

The feature is currently gone. It may be gone for good.

In December 2023 only 10 books were downloaded in Speech Central from Bookshare.

It is sad truth that at this volume it is significantly less of a work to download those 10 books manually from the Bookshare website than to rebuild the app logic that does that.

My effort was ongoing for 7 years with no results at the end and that is something that I had to admit.

At this point I suggest everyone interested to ask Bookshare to implement the share functionality in their iOS app. That would be the best solution as that would make it almost as easy as if the support was built-in Speech Central. I did suggest that to them myself but the more people ask them the more likely that they will take that as a priority.

Everything is very nice, but despite the 15 supported languages, it is quite sad that my native language, Turkish, is still not supported.

I would like to share some issues and questions concerning iOS, assuming the thread does not pertain specifically to MacOS.

1. Can we choose only one voice for each language? Can we not make dialect-specific choices? Why are certain voices selected while others are not? Would it not be better to have each language as buttons as in the VoiceOver settings, with each dialect under headings once the button is selected/activated?
2. Can we manually specify the language of the text we would like to listen to? If not, this would be great.
3. It's awesome as it is and can be improved even further, but I would like to know how the emotion thing works if it's not a secret. I might even help improve it if that is feasible.
4. The pause/resume and speed adjustment features sometimes do not work as expected. When I use the magic tap (two-finger double-tap) to pause reading, it may work and then resume from the beginning of the sentence, work but resume exactly from where I paused it, or may not work at all. What could be the reason for that? As for the speed adjustment button, do we double-tap on it, or swipe up or down? I would love to be able to specify a speech rate for each language, for I can listen to content in several languages or dialects, but it is much easier for me to understand some in much higher rates, while this is not the case for certain others. So I would like to specify standard and high speech rates for every language and even text/book/article/whatever I listen to.
5. I can't adjust the pitch in the voice settings by double-tapping on the slider and holding my finger and slowly flicking up or down/right or left. I have to swipe up or down with one finger, which increases or decreases the pitch by no less than 10. I want to make smaller increments/decrements.

1. If you refer to the Mac it doesn't have specific Multiselect mode as the Mac table selection tools mostly resolve this, so you can select one or more voices and selecting a new voice may or may not replace previous selection depending on how it is performed (I am not sure how this is done with VoiceOver but it should work similar to how you select files in Finder).

If you refer to the iOS (which seems to be the case), the default selection mode works exactly like you have described - it will allow you to select one voice for each language in broad sense (including all regional variants). If you switch to Multiselect you can select more than one voice for each language and the app will use randomly some of those voices on the document.

2) When you change the voice manually during listening the app remembers this as its new language.

3) Currently I don't disclose much regarding the Emulate book narration mode. But I do plan to allow the customization of the effect in the future so it will be a bit more clear on how it works.

4) Both of the described cases (aside from the one where it doesn't work) are by design. If you continue fairly shortly after the pause it will play from where you have stopped as it is reasonable to expect that it feels natural. If it is a longer period of time at the best it won't feel natural and at worst one might not even remember anything he heard so the sentence is restarted.

5) The only thing that the app does is that it adjusts the increment for the action of swiping. I am not sure if it may interfere with something, but it is good to know to test it. If it does interfere it is Apple's bug in its SwiftUI (which is a fairly new thing so not too many apps do use it yet).

No, I was referring to the iPhone as stated in the beginning of my post. So is what multi-select mode does let us pick multiple voices per language? But why and how does the app then pick one of them at random? Would it not be better to be able to specify the dialect for each book or article we listen to? One might wish to hear something written in British English from a British voice, and something written in US English from a US voice. Likewise, other languages also have various dialects, and these dialects may differ in terms of spelling, so there are cases where text read out by a voice speaking a certain dialect, becomes much less intelligible. By the way, I had only one voice under the Australian and Indian dialects for English, but VoiceOver would not say "Selected" when I would double-tap on them and then refocus the cursor on them. One more problem is that I have the compact and premium versions of Maged (an Arabic voice) downloaded to my device, but the enhanced version also appears as an option. You might also consider defining eSpeak-NG as a "robot" so that its variants/voices are also excluded if we enable the option to exclude robots. Alternatively, you may do that only for the klatt/robosoft/NV Speech Player voices (Adam, Benjamin, Caleb and David). You will get what I mean if you download the app or have already used it. As for the narration mode, you may also contact me privately, as I am willing to help it become even better. One thing I can say for now, is, as far as I understand, it raises the pitch when it encounters sentences ending in exclamation marks or phrases in quotes; however, this does not happen for some sentences, even if they are part of the same piece/block of quoted text. The pause/resume feature works quite reasonably according to your explanation, but it could still be more customizable. The user may choose to have speech resume from the beginning of the sentence, or from exactly where it was paused, in any case. I am sorry to know you may not have much to do to fix the speed adjustment bug, but you can at least try to reproduce it and trace the source.

I can send you sound effects I recorded and synthesized myself and CC0/public domain sounds from other sources to play when moving to the next or previous page or paragraph and even sentence, to denote omitted content like footnotes, when deleting a book or article and in various other cases. One more suggestion I have is that the user may be given the option to not delete the things (s)he trashes immediately and instead store them in a Recently Deleted folder as in certain other apps like Files, Photos and Voice Memos. You might also consider inserting longer pauses for paragraphs and shorter pauses between sentences, and let the user customize the durations. You might even make the pauses speed-dependent, where every pause value shall be divided by the speech rate and increase or decrease as the user changes the speech rate.

I have implicitly mentioned one of the guiding principles of architecting the app, but let's say that there are two principles:

1. unless there is some physical limitation there is usually a creative way to make the app work for everyone and that is the solution that I am looking for

2. as long as such solution is impossible/considered to be impossible at the moment, the other principle is the principle of overall usefulness - every design choice impacts various groups of people in positive and negative ways. This has a lot of nuances like that those groups of people may be of unproportional size and also that the amount of those positive and negative effects might be unproportional. Then you consider which solution is overall better for the app audience. Which obviously may mean that it may be significantly worse for some users.

When the app has started 10 years ago it has actually tracked item languages per their regional variant. Some of the content may include this definition in its source and then you would get for example BBC with the UK accent. However it turned out that majority of users were very unsatisfied with this. So I certainly do respect your idea and find it valid as the app was working like that. But that is minority use case, unfortunately.

Further, there is some sweet spot between number of settings, their organization and users ability to actually set what he wants to set. Certainly everything can be customizable but with too many options, the users are going to be lost. I have already considered the option to customize on how the pause would work before, but the number of people interested in that doesn't justify added complexity of the app for other users (to be honest thus far in 10 years and many thousands of feedback pieces received no one was interested in this particular thing).

Considering the voice selection feedback bug I could reproduce that currently doesn't work as expected with VoiceOver. This is again the bug in SwiftUI. As something that Apple uses more widely starting from this year SwiftUI is still likely to have some minor problems with VoiceOver. I did notice one more SwiftUI bug that wasn't reported, and it was that Edit mode in the app didn't allow selections. I could find some alternative controls that eventually worked, so it works properly in the current versions. I have some ideas for alternative controls here too, but you can't say whether it is going to work until you try it.

Regarding the narration mode it is really a fairly complex analysis of the text that spreads to several thousands lines of code. Some of the more dramatic effects are easier to spot, but they represent really just maybe 1% of parameters being tracked. As such nearly all if not all suggestions made in the previous post regarding that have already been implemented.

One thing to note is that this particular area shouldn't come with trade-offs across user base, so as long as it is technically feasible suggestions are to be implemented and I look forward to them. Please use the Contact Support in the app or whatever channel you prefer (you can even post them in public here or via DM or whatever is the easiest way for you) and send them. I do have currently plan for two-three new features in this mode that will further enhance realism in some cases but they do require a bit more of fundamental changes so they may not come immediately.

I failed to address one suggestion - Recently Deleted feature exists in the app since 2016. In the current version on the iPhone it is in the Extras section.

Okay, mine are just suggestions, and I know they're not ignored. One thing is, you may give the user the choice to pick a voice for a certain language regardless of the dialect or to choose different voices for each dialect. The app already has a heading for the language and then headings under that for different dialects. So if the user selects a voice for the language but does not select any voices for any dialects, then the generic voice should be used for any content in that particular language. Also, let me remind my suggestion that we should be able to manually specify the language and dialect of books and articles, and both automatic and manual language detection features should work within one single book or article. The user might even be given the option to specify the detection unit, like "document", "page", "paragraph", "sentence" and even "word". The user should also be able to specify in what languages the whole book or article contains text, and then leave the rest to the AI or code that detects the language. For instance, I might choose quite a small unit (word) for language detection, and then, in order to eliminate errors, I might instruct Speech Central to look for words in English and Turkish only in that particular book, if I am sure it has nothing in any other language. You may also implement the pause options and gather all the advanced settings in the advanced settings section. This way, only those wishing to mess around with them can change them and they do not confuse the other users preferring a more simplistic approach. I understand your situation and that certain features require much more time and effort to be implemented, but my suggestions apply to cases where you can actually implement a certain feature without any major difficulties but are concerned about the user experience being impacted adversely.

I could reproduce the problem regarding the bookmarks not being visible on load on the Mac and it will work in the next build that is already submitted to Apple.

I think that I was able to reproduce the other problems though they appear on a very specific content - only on some very specific web pages. That is also fixed in the next version, assuming that that is it.

Thanks for taking a note on those issues!

So this wasn't happening a few days ago, but today when I started reading a book... y speech rate is 2.0, at about 285 words per minute. I'm using the elequense reed, although tested this with other voices such as alex and... same bug. So, once you press play, the book starts reading as normal, but after a sentance or 2, the rate goes from your set rate to... Extremely; and I mean extremely... slow. Even if you say, change your rate, like increase your rate to say, 2.1 and then decrease it back to 2.0, this still happens. I reseantly downloaded the apple watch speech central app but, I don't think this has anything to do with it; although I cannot seem to get books to read in apple watch either. I open a book and it just says loading, with apparently nothing happening. As far as the bug I mentioned, it’s happening on ios, will test on mac to see if it does but, here's a screen recording. Note: I tried this with multiple books. Books which never had this problem before.
https://u.pcloud.link/publink/show?code=XZEwm50Z6OfBDqVM1CXCKDeIYNJ8zBhC18Qk
Edit: Ok, ust tested this on mac os, and on mac, latest speech central version, this bug isn't present whatsoever, it just seems to be on ios.,

As with any other software work, all reproducible bugs are fixed, all non-reproducible bugs are ignored until they become reproducible.

So the first bug mentioned, bookmarks not loading while opening of the new Items with bookmarks sidebar enabled by default had all necessary details to be reproduced at first and while it wasn't confirmed above, I am 100% sure that this is fixed now. I have tested by downloading the app from the App Store.

The other comment that bookmarking doesn't work is an example of the report that lacks details to be acted upon. After careful examination I have found that it doesn't work on one specific type of content as said before. And it was a matter of luck that I actually found that. It is certainly possible that there is some other type of content that may trigger the problem, but that doesn't make this situation actionable.

Regarding the Eloquence voices, I couldn't reproduce this and I would suggest the device restart to ensure that it is not something temporary. Either way if you notice this only on Eloquence voices then it is a matter of their software bug. Which shouldn't come at a surprise, they should come with label "use at your own risk" just like some other voice software from Apple (e.g. German voices). As you are probably aware one other app doesn't even allow you to use them, I would assume it is because they had some horrible bugs in the last year.

Finally, regarding Apple Watch, you should be able to open any book on Apple Watch. However please note that, just like iOS Apple Watch stops any app work when the screen is dimmed and some longer books may not be opened in the period before the screen dims, especially on lower end devices and longer books. However there will be notable improvements in this in the next month or two, some improvements should be already made in the latest version of the app published yesterday.

I kindly remind you to respond to me.

@Labsii, This isn't just a eloquence thing, per this screen recording I just made,
https://u.pcloud.link/publink/show?code=XZ4Ij50Z7bieTa90RiYboJenCud9BHUPHWfk
Also, I tried restarting my phone and... bug still persisting.
Edit: Checked the appstore for an update to speech central, I downloaded the update, but this issue is still persisting.
I figured out the bug! Check this out.
https://u.pcloud.link/publink/show?code=XZVNj50ZQzylAhIO8NkLieGibQ18Q4Qw2K2y
Long story short, to reproduce this,
Go to settings, advanced, and set the speed unit to wpm, constant, then let the voice to keep reading the book. As you can see from the screen recording, this bug only happens if you have the speed unit set to wpm, constant. If you set it to any other setting, ratio, wpm average etc; everything works fine.

I cannot say what is the cause of non-reproducible bugs. I can say that each of them certainly falls in some of those categories:

1. temporary or permanent software corruption on the device (which then can be in the operating system part or app part)
2. some specific document content required to reproduce the bug
3. some specific setting (or combination of settings) required to reproduce the bug (which then can again be in the operating system part or app part)

I can make educated guesses on the likely cause but I could say for sure only if I had the device and in most of the cases that is impossible.

So for the bookmarking problem I do expect it to be reproducible in some context and as such video might help to better understand the context.

For the voice problem that looks like the operating system problem. As this is happening only on higher speeds, you may try to disable Speed Reading Optimization in Speech Advanced settings to rule out if this interferes in any way. But I did try both cases and couldn't reproduce.

Regarding suggestions I try to address those when I have time. I welcome any suggestions and read all of them, but deep discussion of them is not always possible. At the moment I do have several higher priority tasks that I need to complete including some business administration tasks that come with high penalties if they are not executed in proper timeline...

The number of threads on Speech Central is probably literally in the tens, which may make relevant information more time-consuming to find. I therefore wanted to post the problem here instead of making yet another one of that plethora of threads. So when I begin listening to a book and then double-tap on the standard/high speed slider, the current sentence is read at that high speed but then I hear something like "rate=2 1 1" and the speech rate reverts to the standard value. Repeating this causes the exact same problem to occur only a sentence later. I don't know whether this has ever mentioned before for the reasons I mentioned above, but this issue appears to be quite urgent.