darn it I wish I didn't like the 22 kHZ quality Eloquence so much, it's all I ever use. On iOS, I check the "higher quality" box. I need it and want it, it'll be a sad day when I'm forced to use 11 kHZ Eloquence. One reason I don't use JAWS much despite it having Split Braille. 22 kHZ Eloquence is the bomb, it's the beast, the bee's knees. It sounds so much better and not like an AM radio Eloquence (11 kHZ)
I also hear this odd transition from some voiced to unvoiced sounds in 22 kHz. All I know is, I literally can't deal with it for more than about 10 seconds. The longest I ever lasted was a couple of minutes.
@BorrisInABox I find the same with natural voices--punctuation is a period, sometimes a comma, sometimes a question mark, and that's it. I hate that.
I like ESpeak because it's the best at figuring out how to say odd compound words, which I run into a lot as a coder. It's clearer, too, where I find Eloquence can mumble. I also don't like the way the pitch drops if it doesn't encounter the right punctuation.
Overall, I'm glad there are so many voice options.
So glad I've never liked Eloquence 22K - LOL! To me it sounds like a decrepit, old person attempting to read stuff, and it brings out the harshness of the voice, something 11k properly hides. @BorrisInABox
Espeak sounds like it uses 22K to me, like there isn't an AM Radio drop down when you switch to it. I wonder if a lot of blind people just got used to the dampened quality of 11K voices and now, when a higher quality 22K robot voice is offered to them, they find it it too harsh? Hmm. Very interesting to me.
One of the things that bugs me the most about eSpeak is those pre-recorded consonant sounds, especially the cl/kl variant, like in the word click. It bugs me about as much as this one Alex breathing sample with a mouth click in the middle of it.
@amir @BorrisInABox It wasn't an instant transition for me, but once it clicked, I could never go back. I use other voices on my Apple devices. I've tried RH Voice (I like SLT best so far), Eloquence, Dectalk, OneCore, and others. Nothing matches the clarity, handling of odd words, and consistency of ESpeak.
@alexhall Using ESpeak is something, and liking it this much is something else. I have to use it for Persian as I have no other choice if you don't consider a terribly-sounding Vocalizer Persian voice, but, TBH, really hate it. ESpeak is much more robotic than Eloquence at least as far as US English is concerned. I've never been able to get used to it. @BorrisInABox @Tamasg
@BorrisInABox @amir I've never noticed that ESpeak uses recorded sounds. I thought it was all, what's the word, Klatt-based? But yes, I've never liked that one breathing noise Alex uses. I don't know if it's a mouth click or what. It always sounded like a weird hitch in the audio, like my speaker was about to go all staticky on me but it never did.
@BorrisInABox @alexhall @amir LOL wait the Alex voice has a mouth click in the middle of a breath? Now I'm going to be looking out for this on my work Mac, thanks for pointing it out. :D With Espeak for me at fast rates the letters N and M feel like they're slower than the rest of speech, which is what I notice. So like, 19 of 29 with fast Espeak, Max voice (default) and American English, it almost sounds like it slows a split hair down on the N, and I instantly notice it in the middle of other words as it speaks.
@JamminJerry @amir @BorrisInABox huh so there's a group of fellow 16K and 22K Eloquence lovers out there? Not sure. So far I've only come across haters of it. I'm amused that to my ears it sounds clearer, cleaner, not as sharp, like a soft blanket wrapping me in the soothing sounds of roboticness
@amir @BorrisInABox I don't even think the eloquence at 16K sounds quite as good as the 11K does. maybe it is just me so used to the 11K one, but I am playing with it on android, and set it to 16K, as there are people that say it makes it sound better. I personally don't think so.
@alexhall @BorrisInABox Absolutely. Eloquence just sounds pleasant to me, don't know how else to say it. Just as good pronunciation as the pete voice from double talk or paul from dectalk.
@alexhall I need to try those Blastbay voices. Perhaps they really will be the savior to all these problems, sound more natural, do punctuation right, sound consistent as a 22K voice to all, decently low lag, all of it , magically solved by one engine. I'm really, really trying to find something. Truthfully every day I do feel like the end of the road for Eloquence is drawing nearer and nearer. Sure it'll stick around for 3, 5 more years even, longer perhaps, but the architecture around it will change so much that it just won't be able to keep up with it. Already, if you run Eloquence on an ARM mac, you hear pops and clicks in the middle of words. It's annoying as fuck, because the ARM conversion somehow from X86 changes enough of the speech processing that pops and clicks appear. @amir @BorrisInABox
@alexhall I can give most of it to you. But as someone who works on Eloquence pronunciations and compares TTS engines, I can confidently say that eSpeak is relatively bad at handling odd words. Sometimes a bit better than, and sometimes worse than Eloquence. It doesn't respect proper English stress patterns with longer words, so sometimes you may wonder which syllable has major stress with eSpeak. @BorrisInABox @Tamasg
@amir @BorrisInABox Odd words for me often include coding things, like variable or function names. I've found that when Eloquence slurs everything together and tries to say it all as one word, ESpeak is much more likely to break it down into multiple, recognizeable words.
@JamminJerry @amir Yep. the 8 kHz isn't just a straight sampling rate conversion. The upper formants are shifted down. I have actually used that in some specialized applications.
@BorrisInABox @JamminJerry @amir weirdly on 8K Eloquence I hear a slight whistling sound, especially on the letters s and X, or the CH sounds (like much, crutch, ETC.) Pretty sure it would be the "tel" voice as IBMTTS called it, and what NFB Newsline used for quite awhile before later using standard 11K. To get the 22K version, it's more involved since it needs another DLL that's version 6.4 or newer of IBMTTS, and I think if you already dislike the 16K the 22 probably won't improve much for you anyway LOL.
I just found this. Read the word Skill over and over again with Espeak. OMG. Yes. Pre-recorded. You're absolutely right. Skill, skill, skill. @alexhall @amir
@amir @BorrisInABox I don't know about the 22K ones, but I have seen I think two people say they like the 16K version as apose to the 11K version. I just looked, and my IBM one in NVDA only has 11K, and 8K of all things. it doesn't even have 16K, let alone 22K.
@GamingWithEars That one I can actually see a bit. I think for me it took some time to get used to it, but now going back to 11K actually sounds worse after forcing myself to just get over the S's thing and use 22K for awhile, it's so odd how the brain just can accept it if it's consistently the same.
OK, that's it, DECTalk is the winner then. Clearly. DecTalk. 11K version sounds not quite as AM quality just a little bit, 22K variant is over-bassed and too thick, but hey, it's the OG. Klatt at its finest. LOL. So we have: Bestspeech, ESpeak, DECTalk, Eloquence, Orpheus as our formant voices. Not sure if I'm missing one there. I honestly do wish we had more options, wish I could revive the Doubletalk voice under Windows somehow but it was too monotone for many and there's no DLL of it that exists anyway. @alexhall @BorrisInABox
@alexhall @BorrisInABox ESpeak literally makes my ears hurt after a while. Such a grating voice. Some variants are a bit better than the default, but I even prefer RHVoice over Espeak!
@alexhall @amir @BorrisInABox Thye really do check all those boxes, and the sub20ms latency is only going to go down further still as further optimizations get made. And it's coming to Android next.
@alexhall @amir @BorrisInABox I like the Blastbay voices, but don't use them with NVDA all the time as they aren't as responsive as Eloquence or Vocalizer, I need to check again to see if they do punctuation inflection and let you know.
Jumping in here because I'm weird and use a voice that about 0.0001% of people use. Orpheus. Can't work well without it. It's my Eloquence. Without access to that, I'd probably still not be an NVDA user at all. @alexchapman @Tamasg @alexhall @amir @BorrisInABox
@FreakyFwoof heh, I was very sad, my Orpheus broke after getting the last NVDA, never found a newer add-on version for it that fixed it. So it now sits there in my synth dialog all abandoned and broken, ha. @alexchapman @alexhall @amir @BorrisInABox
@FreakyFwoof @alexhall @amir @BorrisInABox I've tried using that and after a few minutes, it does my head in IDK why. And I don't know if that is gonna work with 64Bit NVDA.
@alexchapman @alexhall @amir @BorrisInABox Nope, mine still works unless you're talking about the 64-bit version. It's working fine here in 2025 builds.
@alexchapman @alexhall @amir @BorrisInABox That's me and Eloquence. I'm a snobby Brit and cannot stomach US voices as my daily driver, and UK eloquence is well.... Let me not pollute the air with my filth.
@FreakyFwoof Oooh, I had 0.21, not 0.22 driver, so now I updated it and it works. I do like this voice a lot. For me honestly it might be closest to replacing Eloq, even if it is 8K, but at least it sounds consistently good and very lag-free. Consanants at 8K don't sound muffled or like a filter is before his mouth. However, Orpheus does have a bit of a sound of, man gritting his teeth and talking through them. Like if I had my mouth closed. It's about the one thing I would duck it for. @alexchapman @alexhall @amir @BorrisInABox
@alexchapman @alexhall @amir @BorrisInABox Well, no. In the advanced tab it says 22 K, you can try changing it to 11 if you really want but I wouldn't advise it.
@alexchapman yeah was surprised but both the Light and regular versions there got an update this April, I just never re-checked to see if the file timestamps changed until now. @FreakyFwoof @alexhall @amir @BorrisInABox
@FreakyFwoof @alexchapman @alexhall @amir @BorrisInABox i use it on occasion. I actually use a voice that no one wants to use anymore, or people can't get as easily anymore. Acapela Heather! Even on my android phone, used to end up using Dectalk as a daily driver because google TTS, is a nono, and ESpeak, I don't really mind it, but, it changes languages a bit too often and that's a glitch I'm not gonna deal with. Me personally though, I prefer Dectalk over Eloquence, because it has more inflection. But if for whatever reason, multi TTS won't work, then i'm just going to use Dectalk, when I get a Pixel that is, because Google TTS is, no. Not really reliable.
I got Jake to update the addon with a dictionary that myself and Derek C maintain, but we've been adding to it since the addon was last updated, so I'd advise loading this up if you wish to. Also for more inflection, under the voices tab, select the voice you're using, click Edit, find Intonation and knock it up to 100%.
@TomGrant91 it requires the IBMeci.dll file, version 6.6, to even show up, and most people are running the standard 6.1 DLL. Quite a bit of fiddling to get it working right, I host it at eurpod.com/eloq66.zip people just have to install datajake.braillescreen.net/tts… for Viavoice light, then replace files in program files from the zip with the ones in there, pointing IBMTTS to it afterwards. May or may not work so have another synth ready if you decide to try it, but most people hate it from what I'm seeing.
@FreakyFwoof wow it really is 22K, changing it to 11K makes it sound like a chipmunk because then you have to edit the player to also use the correct sampling rate inside the _init.py file. Very interesing, would not have thought they defaulted it to 22K, wow. Another reason to maybe switch to it, LOL. @alexchapman @alexhall @amir @BorrisInABox
@FreakyFwoof @alexhall @amir @BorrisInABox I just tried installing it and got the following: Error dialog Some operations on add-ons failed. See the log file for more details. The following add-on failed to be installed: orpheus.
@alexchapman @alexhall @amir @BorrisInABox 64-bit or 2025.x? If so, did you install the most current version? Doesn't yet work with 64-bit but I'm hoping someone will make that a thing.
@alexchapman that's odd, you could try seeing if it made the .tmp folder for the add-on, which it does when it extracts. If so you know it got passed uncompressing stage. Another option is to just manually create an orpheus folder in your add-ons directory and uncrompess there. Brutal but works when I get that error. @FreakyFwoof @alexhall @amir @BorrisInABox
@alexhall @amir @BorrisInABox The Blastbay voices are nice and responsive but no, they don't solve all those problems. If anything they add some new ones.
I have one, Libby, but it lags randomly and isn't great with punctuation. I loaded it up to test it, and lost all speech in NVDA after 20 seconds. I had to restart NVDA, and when I did, it came up using ESpeak again. Libby doesn't seem reliable enough for daily use.
@Scott yeah just tried them. They still treat exclamation marks and period marks the same. Question marks do get pronounced right, but that really only puts at the point of where Vocalizer is in terms of punctuation handling. A bit disappointed by them, but glad I had the trial. @alexhall @amir @BorrisInABox
@alexchapman It really is not a bad voice. It also doesn't inflect on exclamation marks versus periods, same reading. But I tried the same passages with Eloquence, and it's the same! I just can't believe people read stuff without ever knowing there's an exclamation at the end! Eloquence handles that slightly better at long passages with exclamation marks, but to think that we're so ignorant of them because of our TTS engines is a bit unreal to me.@TomGrant91 @mckensie @FreakyFwoof @alexhall @amir @BorrisInABox
@alexchapman @TomGrant91 @mckensie @alexhall @amir @BorrisInABox It does with intonation set to 100, but it's more a feeling. I think I can usually tell when there's a period or exclamation but I've used it for 20 something years.
NVDA >Tools >Configure Orpheus. Control tab twice, set the language to UK, then either tab to Exceptions Editor or do alt e. Follow that up with File >Import, or Alt F, I. @alexchapman @Tamasg @TomGrant91 @mckensie @alexhall @amir @BorrisInABox
@mckensie Not yet. I am trying the latest copy of the voices on there, it reads them the same. It's crazy how I'm having a hard time finding a voice that handles this well. The unicorn voice that does exclamations good. At this point I wouldn't mind if it yelled loudly when it encountered one, like as though fireants were stinging the voice's butt. @alexchapman @TomGrant91 @FreakyFwoof @alexhall @amir @BorrisInABox
@FreakyFwoof do you know any way to reduce pauses? Like when it says buttons. "new.... button." huge pause there. If there's a way I'd love to improve the add-on is to remove those, but no idea if even Orpheus would support such a thing. @Scott @alexchapman @TomGrant91 @mckensie @alexhall @amir @BorrisInABox
@Scott @alexchapman @TomGrant91 @mckensie @alexhall @amir @BorrisInABox Paperback for example, has an Open button with three dots and it pauses before telling you the control. I'm always annoyed when a voice doesn't pause in that space, probably because that's what I got used to.
@FreakyFwoof oh that's wild. For me having "shorten pauses" on with Eloquence did quite the opposite. Wow. Well, I might craft it up. Inside "edit" I only see: age, gender, headsize, intonation, pitch, voicing, and voicing pulse. No phrase pause setting there sadly. But I wonder if I could just do a rule where it throws away extra punctuation marks and makes them just one instead, at the very least it's a hack. @Scott @alexchapman @TomGrant91 @mckensie @alexhall @amir @BorrisInABox
@TomGrant91 @mckensie @FreakyFwoof @alexhall @amir @BorrisInABox I finally got Orpheus working, and am trying Andre's suggestion of 100% intonation and its weird lmao, oh and the monotone way it reads letters as you type, I'm so used to individual characters at least being read out with some inflection lol.
@alexchapman @TomGrant91 @FreakyFwoof @alexhall @amir @BorrisInABox Speaking of TTS engines and Exclamation marks, I don't know if you guys knew this, and perhaps it might be up now, but Philip is planning to update Blastbay TTS to make so that it handles exclamations and such. I don't know if he has added that in but maybe we should check. I'm almost tempted to check his page.
@TomGrant91 @alexchapman @FreakyFwoof @alexhall @amir @BorrisInABox I wasn't exposed to Orphius in my early days of technology, so when I hear about it I was like, what even is this? Guess it partly comes from me being from the US lol
@mckensie @alexchapman @FreakyFwoof @alexhall @amir @BorrisInABox I think when I was younger, I assumed every piece of talking tech had an american voice and thought supernova would. Talking LM-6000SE dictionary, Bestspeech US. Keynote gold using a very old JAWS on a laptop for touch type practice using megadots. From then on to Dolphin Supernova V5 and Orpheus, it seemed weird that the first language is UK, expecting US English.
@mckensie @alexchapman @FreakyFwoof @alexhall @amir @BorrisInABox with ya on that one! Despite the roboticness of orpheus, I could use it daily, the UK accent is much much better than eloquence's. Orpheus was the rfirst synth I used straight with the UK voice in new college worcester and it became their version of Eloquence to me there.
@FreakyFwoof @Scott @TomGrant91 @mckensie @alexhall @amir @BorrisInABox Oh wow, I also noticed the dictionary editor is basically a list, and its easy to add new entries, so if I notice it saying something odd I can try to correct it.
@luiscarlosgonzalez requires the IBMeci.dll file, version 6.6, to even show up, and most people are running the standard 6.1 DLL. Quite a bit of fiddling to get it working right, I host it at eurpod.com/eloq66.zip people just have to install datajake.braillescreen.net/tts… for Viavoice light, then replace files in program files from the zip with the ones in there, pointing IBMTTS to it afterwards. May or may not work so have another synth ready if you decide to try it, but most people hate it from what I'm seeing. You can always do synth dialog and press W for Windows OneCoreVoices to get back if it fails. @FreakyFwoof @Scott @alexchapman @TomGrant91 @mckensie @alexhall @amir @BorrisInABox
Am I the only person who likes UK Eloquence? In fact that's my normmal voice. I sometimes even read Spanish with it if I'm too lazy to change the language.
Today, we fined X for non-compliance with transparency obligations under the Digital Services Act (DSA).
We're holding X accountable for: 🔹Deceptive design of its ‘blue checkmark’ 🔹Lack of transparency of its advertising repository 🔹Failure to provide access to public data for researchers
So helping my wife make a 40 question google form. It looks like splitting questions in to sections is not accessible since you can't move the sections after they are added. Same for questions, I added them in the order we wanted but they are now in a random order and cannot figure out how to move them. I assume it is a drag and drop them. The shuffle order for questions checkbox settings has always been unchecked? Is there a better/more accessible platform for form creation out there?
Syncthing-Fork für Android: Original-Maintainer verschwunden, Repo & Signaturschlüssel landeten bei neuem, kaum bekanntem Account. Kein bestätigter Schadcode, aber Supply-Chain-Risiko. Empfehlung: Auto-Updates stoppen und auf letzte als vertrauenswürdig geltende F-Droid-Version 2.0.11.2 einfrieren, bis die Lage geklärt ist.
Welche F-Droid-App (bzw. quelloffene App) kennt ihr, die großartig ist, aber kaum jemand auf dem Schirm hat? Exotisch, nischig, unterschätzt – egal, Hauptsache spannend. Stellt hier eure Geheimtipps vor! 👇
the last 3 releases (2.2.4 to 2.2.6) have no APK attached – hence, IzzyOnDroid is stuck at 2.2.3 (the last release that had an APK). Could you please attach them again? Starting with the currently ...
This year brought great improvements to #accessibility in GNOME, including:
• Accessibility from the start on the login screen • Full accessibility of GNOME Web • Tons of improvements to Calendar • Screen reader integration for Notifications • Configure #Orca screen reader from Settings • GTK apps integrated w/Windows & macOS accessibility
Help us reach 1,500 #FriendsOfGNOME so we can focus on accessibility even more in 2026!
I saw y’all introduced an end-of-year donation nag similar to @kde! (I say this endearingly and happily; FOSS needs to be okay with asking for user support!) Perhaps it was already there, but it seemed new. Do you know at this stage if it’s provided a significant boost in donations, similar to what KDE experienced last year?
@moshimotsu we introduced the desktop notification in the latest release (GNOME 49), so only folks running the very latest will see it. But yes, we started with zero #FriendsOfGNOME through the new donate.gnome.org and are now at 777 recurring donors!
We don’t currently track the source of donations (e.g. whether someone is coming from the notification versus social media or the GNOME.org site), so it’s hard to attribute it all to the desktop notification, but we consider it successful!
@moshimotsu it is also patched out or disabled in some popular Linux distributions, which can affect its success; we’re chatting with folks from those distros to see if we can work something out together.
That’s unfortunate, and feels a little counterproductive. As for the “source” of the donation, I’m not sure KDE tracks those either, but the perception of success was largely brought on from seeing an unprecedented spike in user donations right about the time of the nag. Yes, correlation is not causation, but Occam’s razor may apply!
@moshimotsu To the crowd, this is deeply saddening because the reminder is the only way we can reach folks that do not participate as an enthusiast. For instance, the parent that had their computer installed by one of their kids. They are not participating in social media or in spaces that enthusiasts are. A gentle reminder that this is a community and we need their support is perfectly reasonable.
@MikeGorse Orca configuration, within Settings, makes sense; glad to see it! I haven't done with GNOME Web lately, but should have a look. Awaiting latest changes, to Calendar accessibility, in GNOME OS Nightly. Good Work!
@wx1g @MikeGorse we were a bit constrained by the character limit; to clarify, you can launch the Orca settings from within GNOME Settings, now. It’s not currently configurable within the Settings app itself.
I will try out your Desktop GUI.......not sure I like the way it is laid out since GNOME 2.x but thank GNU for your hard work. I wish I could program to help support free software but I am not that gifted. ❤
if you want to build a secure system, you have to assume the dumbest (who isn't dumb, just not a nerd like you), most naive (who isn't naive, just doesn't know tech like you do), most helpless (who isn't helpless, just doesn't understand tech terms, especially in terrible documentation), most impatient (who isn't impatient, just doesn't want to study computers for 15 years) user and design your system around it.
Django 6.0 introduces a built-in background tasks framework in `django.tasks`. But don't expect to phase out Celery, Huey or other preferred solutions just yet.
LibreOffice Conference 2026 in Pordenone, Italy, and LibreOffice Conference 2027 in Gothenburg, Sweden LibreOffice Conference 2026 will be organised by PNlug (Pordenone Linux User Group) and the University of Pordenone in the city of Pordenone, on th…
Latest release of IVPN for Desktop (v.3.15.0) is out with Custom DNS and stability improvements. You can now specify multiple custom DNS servers with optional DoH support.
The Most Unworkable Internet Law in the World: Quebec Opens the Door to Mandating Minimum French Content Quotas for User Generated Content on Social Media michaelgeist.ca/2025/12/the-mo…
The Quebec government has amended its Internet streaming legislation by removing an exemption for social media services, establishing the most unworkable social media regulation in the world with companies required to meet both French language minimu…
2 Cadence Braille Tablets is definitely better than one. If I had to compare the gap, it's like half the width of a word spacing character. So think of the spaces between words, cut it in half, that's your gap size.
Castopod focuses on podcast hosting with ActivityPub, happy to answer any questions! Anyone using Castopod or other solutions - please share your thoughts with @everton137 !
Looking for guidance from my musician friends. I have a Komplete Kontrol A61 keyboard, and think it's far past time I learned to play this thing competently again. I also have a very friendly cat/raccoon/weasel/mythical creature hybrid whose species we aren't really sure of. Said catmonkeyweaselcoon likes to sit on things. As such, the keyboard is currently in its case to keep it from getting clogged with beige fur and litter. The case does a great job of keeping it litter-free, but also does a great job of keeping me from playing.
What do I need to keep this thing safe from our monkeybobcat creature? I see dust covers but our boy is 13 pounds dense, and I'm concerned the keys might get crushed over time. I don't think I want a case, because I'd like to leave it on the stand, though I'm open to a case if it somehow doubled as a stand/cover. Are there hard-shell dust covers? Stands with hard-shell covers that aren't huge? I just have a simple stand right now. Happy to upgrade it if that's the key but I'm having a hard time finding things that will work and figured I'd ask for pointers.
Damn this is a tough one as I don't live with pets, so don't really have any experience here. I know a lot of digital pianos have a cover that you can put over the keys to not only make it look like an actual grand piano, but also protect them, but something like that doesn't really exist for a midi controller to my knowledge.
Actuellement confrontée à un site où : Il y a un overlay accessibilité (oui ça commence mal). Cet overlay crée un lien « aller au contenu ». Ce lien ne mène nulle part. Cet overlay m’oblige à naviguer dans toutes ses options à chaque fois avec Voice Over, même quand je ne l’ouvre pas. Les boutons à l’intérieur sont des liens. Ils ont une icône et un label, qui sont restitués à la suite par voice over. Bref. Arrêtez. #a11y #AccessibilitéNumérique #WebDev
Version 4 of ETI Eloquence for NVDA 64-bit has been released. In addition to fixing indexing issues and updating parts of the documentation, the major new feature in this release is dictionary updates. From NVDA Settings, under the Eloquence dictionaries pane, users can now choose which dictionary they want and update it directly from GitHub. These updates pull the latest daily changes rather than the official monthly releases. The available options are the IBM TTS dictionaries and the Alternative IBM TTS dictionaries. Note: to see the Eloquence dictionaries pane, ETI Eloquence must be selected before opening the NVDA Settings dialog. github.com/fastfinge/eloquence… @fastfinge
As with any open source project, 64-bit eloquence wouldn't be possible without the community behind it. This release brings us the following:
fixes to indexes, and further code simplification (Tha...
The reason I pull from master is because the filename is always the same. It looks like your dictionaries have a different file name for each release, because the .zip file includes the version number, so I can't detect the latest version in an automated way easily. The job for the next release is to schedule the update checks and have an update automatically checkbox. But that requires me learning how NVDA's update checker works, and I wanted to at least get initial support out, plus the indexing fixes.
Yeah same. I'm not exactly into Eloquence and IBM TTS and just use it, so forgive this stupid question. Essencially, I'm using the IBM TTS driver, which allowed me to specify the path to a dll myself, which made is possible to have another addon that ships the dll. How easily will it be possible to migrate from this setup to the 64 bit Eloquence addon?
64 bit eloquence comes with eloquence 6.1 included. So another addon isn't needed. If you want a different version of eloquence, you're out of luck. This is because every single eloquence dll version requires slightly different code to interface with it. I don't have all the different versions, so even if I wanted to, I couldn't test them all. Also, 6.1 is the only version that will work in portable NVDA installations, and has no major bugs. While it doesn't support different sample rates, the extra complexity to bridge every possible eloquence version to 64-bit just isn't worth it for that feature. We did, however, add in support for Korean, Japanese, and Chinese languages.
Correct, it will never be in the store. However, if you were using eci.dll version 6.1 with IBM TTS, it will sound identical. We don't currently have an update system because we're currently telling NVDA lies about the current addon version and what NVDA versions it's compatible with. This is both to force the addon to keep loading on newer NVDA versions without an update, and so that the "allow NVDA to gather usage statistics" setting does not allow NVAccess to gather accurate info on the addon.
@jonathan859 I see. So again forgive my cluelessness and I should know better as a relatively technical experienced user using Eloquence/IBM-TTS for at least 5 years LOL, but as an end user there won't be a noticeable change of how Eloquence 6.1 vs. IBM TTS sounds? Also, I assume using that addon would be quite in a gray zone and hence it's not ever going to be in the addon store? If so, is there an update system?
@Bri Acapela is most likely going to be default, that or Blastbay even. Humanware doesn't use Vocalizer, though part of me kinda wishes they did, as another option.
@mckensie As popular as it may be, I can't stand Vocalizer. And don't like Acapela either. So go figure - LOL! I like Microsoft Natural/ HD voices, and a couple of Siri voices on that side of the pond. @Bri @fastfinge
@Bri As much as I don't wanna admit it, you guys are right. But what's next, are people going to be really wanting, Oh, I don't know, Humanware to put Eloquence on the Victor readers? Not gonna happen. The new braillenote evolve, i can kinda see, but, no.
@mckensie LOL, Eloq on a Victor Reader would cause a dimensional rift. But yeah, I get you. HW’s ecosystem isn’t exactly built for that kind of throwback. Braillenote Evolve, maybe… Victor? Not in this timeline. @Bri @fastfinge
Eloquence64 does it the way jaws does it. Largely because that's the way I want it. We're based on Eloquence Threshold, as I never used the IBM TTS addon for various reasons, so I have no idea what the shorten pauses setting does or how it works. If you want to submit that yourself, though, I accept pull requests. You'll probably need to modify eloquence.py to add the setting, as well as host_eloquence32.py if it doesn't have the existing commands you need.
@garo @jonathan859 The sample rate thing isn't a huge loss for me, but IBMTTS's specific modes related to pause shortening, its anticrash repository, and backquote tag support are game changers. I myself use the shorten pauses at end of text setting, I cannot cope with the default nor all pauses shortened. But I know plenty of people who prefer all pauses shortened, a la JAWS.
I suspect that when NVDA introduces the "secure addons" architecture, that is what will finally kill eloquence. Any kind of secure addons infrastructure is probably going to require code to be signed by the developer and approved by NVDA. As they do not allow eloquence addons in the addon store, and they do not allow us to use there translation systems, there's no way they will sign or approve the addon. So once that happens the only path forward would be forking all of NVDA, and that will never happen. So we can keep it going for now, but its time is limited. They already have addon and corporate security on the roadmap, so I'd give it one or two years. We really, really, really do have to find something other than eloquence eventually. That's why I keep trying every knew alternative when it comes out, and supporting stuff like blastbay with money, even though it's not there yet.
Have you tried adding an unsigned addon to Firefox, lately? You basically have to recompile all of Firefox yourself. The fact that NVDA is open source isn't really protection.
I hope you're right. But even if NVDA never does this, we've still only got maybe five or six years until Microsoft removes 32-bit support from Windows entirely, the way they did for 16-bit apps. Unless eloquence becomes open source, it just can't last forever. Apple has a 64-bit Eloquence, but they're not sharing.
Far as I'm concerned that'd kill NVDA for me. If it doesn't fork, I'll have to do something else (maybe go back to JAWS? Who knows.) That said, I find it unlikely that they'd do that, and if they do, that someone wouldn't trivially produce a version without it.
Ah, the “secure addons” architecture could become a kind of doomsday scenario for Eloquence, at least in its current community-driven form. But even if things move in that direction, I honestly think that layer is going to be aimed primarily at corporate deployments, not everyday users. For regular users, NVDA has always walked a fine line between security and openness, and locking down the addon ecosystem entirely would break a lot of workflows. There are tons of addons, including widely-used ones that aren’t in the official store, that would effectively vanish under a strict signing-only model. The real question is whether NV Access is willing to shoot itself in the foot by cutting off that much of its developer ecosystem. If they impose strict signing only at the enterprise layer, Eloquence would still survive for personal installs. And if they push it globally… well, yeah, that would be the beginning of the end unless someone forks NVDA, which as you said is extremely unlikely. @Bri
NVDA has never really cared about developers that don't put there addons in the store. And as for eloquence, it's actively hostile to us. They may see "eloquence won't run" as a feature, not a bug. We were promised a compatibility layer originally, after all. And that doesn't seem to be happening, at least not for 2026.1. That's why I started the work so early; eloquence is the largest thing that depends on 32-bit, and NVAccess actively wishes we weren't using it, so I'm not surprised. Now they're saying that we will get a 32-bit compatibility layer, but it's going to be part of the "secure addon runtime". To me, this means code signing. And they're not going to sign our code.
The lag that would come from running the eloquence dlls in a 32-bit vm would be...unusable. The 32-bit bridge already introduces a few ms of lag as it is!
Oh, they won't come right out and take it away. They'll just make it increasingly hard for developers until we give up. And justify it with security and copyright. The first step will probably be requiring code signing unless you're running from source and using the developer scratchpad.
Right, but they've gotten legal threats and takedown requests from IBM, Codefactory, and others previously. So they seem to regard eloquence as an existential threat to NVDA. The fact codefactory is bringing eloquence back to Android worries me, actually. I wouldn't be surprised if the current github repo is taken down due to a DMCA request. Including the DLL or not doesn't matter. They own the API specification, and it's not public. So any code that interfaces with eloquence at all technically violates there intellectual property.
@modulux I don't think the plan is to lock down NVDA to that degree. It will probably be impossible to opt into running insecure add-ons on secure screens, and also in the corporate/secure mode. But in other scenarios I'm almost certain there will be a flag.
Even if there isn't, there has to be a way of testing your own code if you're writing an add-on without it being signed/in the store first. Even Apple allows me to run my own apps on my own device with relatively little fuss. So if necessary, we can run Eloquence via whatever mechanism that ends up being. @fastfinge @Bri @amir
@jscholes Right, that timing pretty much gives it away. If Apple had a real 64-bit port, it wouldn’t have waited until Apple Silicon to appear. The fact that it showed up exactly when their translation layer matured makes it almost certain we’re still hearing the same 32-bit build running under emulation. @fastfinge @Bri
I mean we really, really need to do that anyway. The fact that it "still works" is why nobody is putting the three to five million dollars towards this that rewriting a modern eloquence that uses the same dictionaries and language rules would cost. But once the 32-bit emulation layer on Windows is gone, that's pretty much the end for responsive eloquence. And we know about how long that will be, based on when Microsoft removed 16-bit support. Closed-source unmaintained code just doesn't run forever.
I'm sure @datajake1999 will get his hands on the sourcecode one day and start a new Eloquence revolution. Failing that, we can pay a synthesis expert to build a clean room implementation. Everybody is worrying about nothing! /s @fastfinge @Bri
Do we know that Apple has a special 64-bit build of Eloquence? I thought it was generally agreed that they were emulating, as with JAWS on ARM. @Bri @amir
@jscholes Pretty sure it’s the same old 32-bit build running through a bridge layer. All signs point to emulation rather than a native 64-bit port. @fastfinge @Bri
Seems to me that if Apple were able to obtain the source code for Eloquence, modify it accordingly, and rebuild it for 64-bit, they could've done that years ago. Instead, it arrived around the same time as their highly performant, in-house CPUs and emulation layer. @fastfinge @Bri
Same for me. Without eloquence, code that hasn't even been compiled or maintained since 2006, my career in tech is gone. And I find just how fragile that thread is to be alarming. Our hacks to keep it working are only going to be good enough for so long. And then it's all over.
Sorry if it was unclear; I was being sarcastic. I would legitimately need to reconsider my relationship with, and career in, tech if Eloquence became unavailable on my devices. It is critically important to me. @datajake1999 @fastfinge @Bri
@jscholes Sure, fair enough. But the level of concern people have really just reflects how important this actually is. For a lot of us, a TTS engine functions almost like a display. And for people with additional hearing issues, nothing else has that same mid-range clarity Eloquence provides. So yeah, there might be future options or clean-room projects waiting in the wings, but it still makes sense to worry a little. When something this fundamental is at stake, hoping for a backup plan isn’t quite enough. @datajake1999 @fastfinge @Bri
@jscholes @datajake1999 @Bri I am in favor of improving other options, or even creating another, similar, formant-based synth. Perhaps record more voices to RHVoice, or improving ESpeak plus MBrola, etc.
Creating other options is great, the issue is whether they're going to have the same fundamental characteristics: minimal latency, intelligible at high speed, etc.
All of the above have been tried. The closest we've come is with the formant-based synth option; both blastbay and speech player went down that path. But eloquence was created by an entire research lab, with multiple folks with PHDs employed there full time, and lots of government funding. It can't be replicated by a couple guys in a basement somewhere. We actively need to be fundraising several million dollars for this, right now, to avert the impending disaster. But nobody is. I just have to hope I can keep eloquence going until I have enough money to retire. I don't even know if the eloquence sourcecode even exists anymore, and if it does, the special tools for the rule-based languages they compiled almost certainly do not.
It's also possible that in ten-fifteen years, assuming climate change hasn't killed us all by then, we'll be able to hand AI the binary outputs and the compiled code and have it reverse engineer and rewrite the entire thing. This could cost several hundred thousand dollars in compute, but it's the only option I have any realistic hope for.
You're welcome! Happy Friday! Also remember that all of the worlds space agencies are chronically underfunded, so you could be killed by an asteroid strike, at any time, without warning. Also a massive sunstorm could destroy all electronics at any point. Have fun!
Well, I was running open-early-warning-db, radard, and ion-beamctl. Unfortunately my sendmail queue filled up and they all crashed on account of there being no free disk space, and my LibreNotify account was out of credit because they only accepted European bank transfers. @dhamlinmusic @datajake1999 @Bri @amir @clv1
Ewww! You use open early warning db? You should be using xurustuklomp; it’s written in rust and does the same thing! And it uses a new and exciting file format that doesn’t work with any existing tools!
I don't understand why I can't see the settings for Eloq itself. It works, but it also told me that when I installed it from the assets section, I already had that same version installed. It's very odd.
@FreakyFwoof So has this issue been resolved? I suggest switching to Eloquence before opening NVDA's Settings window because if you don't do that, the Eloquence dictionaries pane either may not appear or may pop up at the bottom of the list. If you are on the Categories list, I mean the list of tabs immediately after Apply button in NVDA's Settings window, pressing E should move you to Eloquence dictionaries. @fastfinge
@FreakyFwoof So I just enabled dictionaries and I had to select one from the combo box, hit apply, before I could check for updates and it installed the dictionary.
We talked about it last night, he had no idea, just as I didn't. Also could do with another windows 10 user to test this so I know it's not just my machine.
Maybe. Doesn't affect me much anyway, as eloq is not my primary synth, so it's just a matter of 'Why doesn't it work' more than an actual problem, at least for me.
@serrebi @FreakyFwoof Let me jump in here to let you know that I just set up a portable copy of the alpha of NVDA 2026, and installed the latest eloquence addon, and I was able to switch to eloquence, hit apply, then ok, go back into the NVDA settings, and in the tree view use the letter e to go to eloquence, and see the dictionaries. It looks like the alternate IBM one is selected, but I had to down arrow to the other one then up arrow back to the alternate one, hit apply, then I was able to hit the check for updates, and it installed the dictionary. I am using the latest stable version of windows 11 25H2 if this matters.
@JamminJerry @serrebi @FreakyFwoof That's awesome. The default choice is, as you said, Alternative, as it's smaller and gives users access to something basic. To get the community dictionaries, one should move down the combo box there.
@JamminJerry @serrebi I only ever saw two options, should there be more? Even after checking for updates. Will look again later when I'm back on win 11.
@FreakyFwoof No, only two options. Alternative IBMTTS, which is the basic one and isn't updated much, and IBMTTS (aka Community), which is much larger and is updated almost daily. @JamminJerry @serrebi @fastfinge
@JamminJerry @serrebi That's... Weird. I thought the smaller would be the alternative because, well, alternative. IE, alternate to the thing it ships with. That feels like a developer chose that way of thinking lol
@FreakyFwoof Ah, do you mean Andre, or Louis? Of course, even the larger one has no entries for your first and last names. A couple of years ago it had an entry for Louis, but it was removed as it interfered with some other words. @JamminJerry @serrebi @fastfinge
@FreakyFwoof I see. It's a stroke of bad luck. Louis, with the s sound, is popular in the USA. Even a couple of cities have that. @JamminJerry @serrebi @fastfinge
@jscholes @dhamlinmusic @datajake1999 @Bri Are you sure that creating a new synth or improving one would cost too much money? Couldn't for example a few developers make it, as long as we don't fail to pay some tents of dolars a month?
@jscholes @datajake1999 @Bri I'm using IBMTTS and am on a build from the first half of the year, before the 64-bit changes. In daily use, does this add-on perform similarly? If so, I can update to a 64-bit recent snapshot of NVDA.
@jscholes I see. Honestly, if there hadn’t been a workable way to use Eloquence in NVDA, I never would have switched from JAWS in the first place. I just can’t tolerate eSpeak for English text, no matter how many tweaks I try. For me, Eloquence isn’t a nice extra. It’s the thing that makes reading fast and comfortable enough to actually function. Without it, NVDA simply wouldn’t have been an option. @datajake1999 @fastfinge @Bri
@neurrone What do you mean by customizing it? Dictionary processing can be either on or off, I mean via the deletion of the dictionary files, or you can opt to use either of the two dictionary sets provided. You can also edit the files or the entries yourself, or even create your own pronunciation dictionary. That's a whole host of customization options. @fastfinge @jscholes @datajake1999 @Bri
@neurrone The IBMTTS driver, for NVDA 32-bit, ships with no dictionaries at all, whereas this 64-bit driver installs the smaller, Alternative IBMTTS dictionaries by default. You can remove them manually by going to C:\Users\YourUsername\AppData\Roaming\nvda\addons\Eloquence\synthDrivers\eloquence, and deleting ENUabbr.dic, ENUmain.dic, and ENURoot.dic. @fastfinge That's why I suggested installing no dictionaries by default, and leaving it to users. @jscholes @datajake1999 @Bri
@jscholes @datajake1999 @Bri Is it possible to customize it instead of deleting it entirely? Yeah I agree with having this off by default and letting users choose whether to install them.
@jscholes @datajake1999 @Bri How do I disable whatever changes it makes to the dictionaries? i dislike the way it pronounces some words now, and want it to work like it did in IBMTTS.
There are Eloquence control sequences to shorten the next pause. The IBMTTS driver uses regular expressions to insert those codes at the end of sentences and in other places. @neurrone @amir @datajake1999 @Bri
As with any open source project, 64-bit eloquence wouldn't be possible without the community behind it. This release brings us the following:
fixes to indexes, and further code simplification (Tha...
To localise the settings panel: import `addonHandler`, call `addonHandler.initTranslation()`, and then wrap all user-facing labels (e.g. text of checkboxes) in a call to the magic underscore (`_`) function.
E.g. `_('&Maximum number of history entries (requires NVDA restart to take effect)')`
@jscholes We are not using scons. We are not using the NVDA addon template. We are not using anything even vaguely resembling a sane build system. See build.cmd and build.py and cry. In other words, if it needs to happen, I need to do it myself.
No, but I did forget one thing: you should put a `# Translators: ...` comment on the line above where some user-facing text is used for the first time, explaining what that text is for. E.g.:
`# Translators: the label for the button to update community dictionaries from GitHub`
Ah... well: there is a translation workflow which will pull the translatable strings out of your add-on, put them somewhere for translators to translate, and then push the locale files back into your add-on. What I'm unsure about is whether they'd accept Eloquence as an add-on in that system.
Indeed. Probably easiest to have people clone the repo, pull out the translatable strings into a file using an appropriate tool (or do that yourself and commit it), compile their own translated strings file, and contribute the textual and compiled versions in a PR.
Even though you're not using the add-on template, you can probably reuse the translation-related utilities from it.
@jscholes Yes, this is the way. Unfortunately I only speak one language, so I can't generate a translation, and I can't switch my NVDA to another language to test the translation. So this work is going to have to fall on someone else, I think. I'm not confident enough in myself to release code I can't really test. I make enough silly mistakes as it is, because I'm not actually good at this and just do it anyway.
You don't really need to test the translations themselves; string selection happens via well trodden code paths inside NVDA. As long as the files end up in the right places (which are documented), it will work.
But I agree with your wider point about the rest of it. I would start by at least making sure the code is set up in the ways I described if it isn't already, because there'll be no translation at all without those bits. And then hope someone can come along to fill in the rest.
@jscholes It is not. I'll push the code changes soon. I assume doing that won't break anything if no translations or translation infrastructure exists? Like, what happens if you set your NVDA to Spanish, but no Spanish translation is available? Does the addon just refuse to load?
@jscholes Thanks! This has been helpful. I feel like a sighted developer trying to make something accessible LOL. I have no idea what I'm doing and no way to test it.
Don't tell anyone, but localisation is one reason I don't like working on software for the community. Not because I don't think it's critical; obviously people should have the thing available in their language.
But the common tools are just so... hacky. They feel like they were built with an order of priorities that went developer, translator, user, when I think user and translator should be considered as far more important.
As a result, the developer experience isn't actually that great—relying on magic global state and functions as it does—and translations can end up with anglicised word ordering because the English-speaking developers and tools get in the way.
@jscholes The trouble is, if you make it harder for developers, people like me will never do it at all. Not because we don't care, just because we're stupid and this is a hobby. In an ideal world, translators could somehow translate the text without requiring the developer to do anything but accept a pull request. Again, not because I don't think it's worth the effort, but just because I don't want to write code I can't test.
@jscholes We also aren't technically supposed to host the eloquence dll files on github. That's why the build process is...the way it is. So that when the repo that contains them gets pulled down, we can get them elsewhere with a quick change.
Does anyone know an accessible work time tracking piece of software or web service? And by that I do mean really accessible, not somewhat usable with tons of unlabeled controls and such nonsense. I personally don't, so inclined to craft something on my own, but before I just wanted to make sure it doesn't already exist. #Accessibility #Work #TimeTracking #Freelance
I (sighted) use f-droid.org/packages/com.razee… but after testing it with android talkback I think it misses even the most basic labels for navigation items. I have opened a bug for this but it likely will not be fixed anytime soon
Steps to reproduce Activate Android Talkback and open the app. Tab through the main navigation What happens Screen reader says "Tab 1", "Tab 2", Tab 3", Tab 4" and then continues with "My activity ...
Не смотрел Redmine? Вроде он был неплохой и кажется с time tracking. Правда видел его ещё в начале процесса повсеместной деградации фронтенда, и как оно теперь не знаю. Больше ничего в голову не приходит из открытого, но осознано и не искал. А всё что было хорошего с чем сталкивался в последние годы, было для внутреннего использования.
@Yinshi Redmine знаю очень хорошо и использую. Но мне нужно скорее что-то вроде такого: начал работать, поставил таймер. Закончил работать, оно само часы подсчитало и вписало, куда надо. Такого софта вагон и телега, но доступного не видел.
If you're looking for the absolute most simple, no bells and whistles time tracking and you don't mind using a terminal, TI will get the job done. github.com/sharat87/ti
Should you be fine with iOS or the Apple ecosystem in general for some reason, I recently discovered Timery for Toggl which is a really nice Toggl third-party client. Also Moons by @kslazinski is a nice, local-only time tracker.
local Metro store having tech issues this morning, with all self-checkouts unable to process card payments. staffed cashier lanes seemingly working ok. wonder if it's a local thing or a broader issue
Reading up on the DRAM/NAND prices situation and I've come to a conclusion: expect prices to raise for basically everything in tech. Not just because these things go in almost every piece of tech, but also because the price hikes are so massive that everybody else in the sector will use them as an excuse to justify price increases.
We've already seen this playbook in other sectors. Take the car market in 2020 or the energy market in 2022. Once prices start going up everybody else joins in. Later they can't come down, because once general inflation starts going up then the costs of those companies also go up, so there's no way back.
Also, you know what's funny? AI companies are ostensibly the root of these price increases - even though their demands are clearly unrealistic and overinflated - but their costs will also go up... and they're losing boatloads of money as it is. They're already financially strained, and we can only hope that increased costs are going to drive them into the ground faster.
Everybody knows that the sum of the decimal digits of a multiple of 9 is also a multiple of 9. But did you know that the digit sum of a positive multiple of 999 is always 27 or more?
Over the years, the most unwelcome thing (that I could still find amusing) to discover that as reflection of their respective domestic politics there's always an extant political sentiment even if it doesn't rise to the level of actual separatism, in Acheh, Riau islands, Sumatera, Southern Thailand, Indonesian Kalimantan even Mindanao, of using Malaysia as a discourse totem, the "what if we had joined that federation instead". Unwelcome because a lot of our diplomatic border disputes in this entire southeast asian region are basically frozen and no one really wants to heat them up, and the formation of Malaysia itself has always been treated as a something suspicious post-konfrontasi, that never really went away.
ANYWAY, tropical storm senyar got us all in the end. The floods weren't bad here comparatively but now we're dragged into the Sumatera-Java psychodrama as deaths continue to mount. #tootSEA #Malaysia #Indonesia
I love that Cloudflare periodically provides us with examples to convey to the youngsters what it felt like when you were in the middle of chatting to your friends online in the 1990s, and someone picked up the phone.
I feel like a sighted developer trying to make something accessible LOL. I have no idea what I'm doing and no way to test it.
![Mike [SEC=OFFICIAL]](photo/contact/80/5761?ts=1755299202)
Andre Louis
in reply to NoiseBox • • •