I'm curious, then, how do you decide what uses of generative AI you can personally justify? I promise not to argue about this; I just want to hear your position.
Matrix Conference Ticket Sales, an open source prototype policy server, and registration tokens for MAS. That, and much more happened This Week in Matrix!
:) I just released a new tune called “Far Away Waves”. Lots of guitar and even some piano! :) Give it a listen on whatever services you prefer at this Disktrokid’s Hyperfollow page: distrokid.com/hyperfollow/vict… HOpe it makes your day calmer and brighter! :)
Listen to "The Buccaneers: Season 2 (Apple TV+ Original Series Soundtrack)" available now: https://thebuccaneers.lnk.to/soundtrack The Buccaneers Season 2 is...
is it crazy, though? Have you considered that email clients are really complicated pieces of software that are hard to write and maintain, and that very, very few of them still exist and are maintained to this day? Writing a mail user agent is not something people do "for fun", given the expectations.
📨 The Future of Stalwart: Curious about our Webmail plans and what's next on our roadmap? Read our latest blog for all the exciting details at stalw.art/blog/roadmap #Email #Webmail #AutRustAutNihil
Since we officially started developing Stalwart on September 4th, 2021, we've come a long way in establishing a powerful and versatile open-source mail and collaboration server.
Today saw the introduction of something ntruly worthy of celebration: a modern, 64bit, truly responsive formant-neural sapi voice for Windows. Today also saw the introduction of a recording that is based around the fact that that demo of the voice does actually downsample after a minute of audio. However, it does not do the impending craziness that I put it through before activating the license.
@matt There's an audiogames.net forum thread about it. Part of it is VITS, but he changed it up to output high-res formant parameters instead of trying to go to audio directly, and had some kind of custom klatt implementation that it was being driven by. This, of course, necessitated training them entirely from scratch, but the formant output was used as the feedback. I think he said the data he had though was unfortunately data he had commissioned for statistical parametric synthesis (HTS) so it doesn't have as high inflection as one might want, because HTS really doesn't like it.
Thanks so much for posting about this, I've been in contact with Philip over the last year, and the work he's doing in the TTS space is really something special.
@FreakyFwoof Two hybrid neural-formant voices, Richard and Libby. The voices are neural yet built around a formant architecture, so are near-formant-level responsive and controllable, I.E. with speech rate moddability that doesn't suffer the traditional limitations of natural voices of today, while bringing modern quality to them. And, at long last, some actually good, check that, damn good, 64bit sapi voices with no persistent drm. blastbay.com/voices.php
@FreakyFwoof And the responsiveness is pretty incredible. I think it's sub20ms, which we haven't seen from a SAPI voice in quite some time. it works well especially well with NVDA given the latest update's improvements to the sapi responsiveness
@FreakyFwoof They sound good! As luck would have it, they've been released the very same day I've accidentally hosed NVDA communicating with Sapi5 here. Oops. Probably gonna purchase these once that's fixed.
@Scott @FreakyFwoof Get in touch with Blastbay, Philipp is usually pretty responsive. I figure they are really performant and all, but there is a certain tone in those voices that just doesn't feel good to me. I can't listen to them for a long time i'm afraid.
@ToniBarth @FreakyFwoof Yeah I'm gonna email him to ask about some small EQ tweaks, can suggest a thing that might help smooth speech interrupt too. Really liking how responsive they are man and the price sits just right here, I'm in.
@Scott @ToniBarth @FreakyFwoof Do you guys find, say, Elloquence, not very responsive then? Genuine question lol. Its all I've ever used so maybe I've never noticed. Otherwise what's the use case for new voices?
@bermudianbrit @Scott @ToniBarth @FreakyFwoof Well, some crucial Eloquence license contracts are long dead, Given the amount of Code Factory products that have quietly dropped Eloquence. With that plus the pile inside of a pile inside of a pile of retrofit patches applied to the thing, it truly is only a matter of time until the thing dies for good. Dectalk's situation is too ambiguous to be an all-purpose solutions. Espeak's GPL-ness...
@bermudianbrit @Scott @ToniBarth @FreakyFwoof Yeah, the situation around embeddable voices is really not that great. Cerrence Vocalizer is probably the best thing that exists right now. That or Acapela, which is getting out of the AT / on-device business AFAIK. SVox Pico is technically fully OSS and permissively licensed, but that's only the "inference" code. The voice preparation stuff is either provided as proprietary exes or missing entirely, the documentation around voice training is not that great either.
@miki @bermudianbrit @Scott @ToniBarth @FreakyFwoof No, Acapela still has Humanware. plus some AAC software, for now. Although the only one that's guaranteed on-device at this point is HW, as AAC software can go cloud-native if it wants to. And I"m pretty sure they're doing just enough to kee ptheir embedded library alive, and that's about it.
@bermudianbrit @Scott @ToniBarth @FreakyFwoof Do we know what this thing's actual HW requirements are? It's neural, so I suspect it wouldn't run so well on old / underpowered CPUs with no SIMD extensions.
@miki @bermudianbrit @Scott @FreakyFwoof Its a trained model, but apart from that it doesn't have the highest requirements, pretty sure it'd do just fine on Andre's machine, although its worth a try for sure.
@vick21 @jcsteh @Scott @FreakyFwoof That's by design as it's a formant neural network or something like that. There's a thread on the audiogames forum about the development.
@jcsteh @FreakyFwoof For long strings and some interruptions I can feel a difference between this and OneCore, with OneCore winning speed-wise. Not by a huge amount here, but it's a thing. Shorter strings like OSARA reports, moving around most dialogs etc, they feel super close on my machine. A lot faster than the neural stuff I've been using via... er... methods. For now that's the main thing. I can use one of these on stream and point people at a legit source if they like them/want to get them.
@Scott @FreakyFwoof Good for streaming for sure. I think that responsiveness would drive me wild in daily usage, but I'm mindful that I'm an efficiency nut, so I'm not representative of most people.
@FreakyFwoof @Scott The thing that blows my mind is that NVDA's recent SAPI5 responsiveness improvements come down to only two things: switching to WASAPI (assuming SAPI5 didn't already use WASAPI itself); and stripping silence from the start of utterances. I'm guessing it's mostly the silence stripping that gives the responsiveness boost here. Which raises the question: why the hell do so many synths append useless silence to the start of utterances? That just seems like deliberate self-sabotage.
@vick21 @FreakyFwoof @Scott Just tabbing around in the NVDA Voice settings dialog. It just felt sluggish to me in responsiveness compared to OneCore. Measuring scientifically would require all sorts of set-up I'm not quite willing to do right now. :) Obviously I'm not going to compare with eSpeak; that's an unfair comparison. Note that I have rate for both set fairly high.
Oh, that's nice that they put this pizza box trashcan next to the machine spitting out pizza boxes. The only issue is that the trashcan is apparently put out by the municipality. Because why would the commercial company care about the life cycle of its products.
This is how you Privatize Profits while Commonizing Costs.
"Learn Accessibility" is a course that takes you through the essentials for building accessible websites and web apps. Created for both beginners and advanced users, this course can be taken beginning to end, or used as a reference for specific topics.
@DavidGoldfield thanks for saying that, I feel old lol. Especially as time is going by so fast these days. actually no that's not me on the piano, it's just a piano instrumental I found
Nastal mimořádný okamžik tohoto roku. Poprvé mám týdenní dovolenou a ta začala dnes v 16:00 h. Juchuchu. Mobilní číslo odpojeno, v Outlooku nastavena automatická odpověď....
Provided to YouTube by The Orchard EnterprisesEšče som sa neoženil · Hradišťan · Jiří PavlicaNejkrásnější Moravské Lidové Písně 1 - V Hodoníně Za Vojáčka℗ 19...
John Voorhees wrote a piece for MacStories earlier this week which delves into a new class in Apple’s Speech APIs and compares it to OpenAI’s open source speech-to-text model called Whisper .
Google last week released its Pixel Drop update for June. The monthly software upgrade includes, most notably, the official release of Android 16 for Pixel devices, VIP contacts, accessibility enhancements, and much more.
Through his work, Cameron Cundiff develops tools to make digital spaces more inclusive. So, he created the Accessibility Nerd project to explore the intersection of technology, software engineering, and AI in accessibility.
Our May Mobile Progress Report shows how we're: *getting ready for JMAP in iOS * planing early for TestFlight * coordinating localization for both apps * asking for feedback with the Android beta * improving the account drawer and more
Checklist I have used the search function to see if someone else has already submitted the same feature request. I will describe the problem with as much detail as possible. This issue only contain...
@Lazarus Thanks for the nudge! It looks like we have this in our triage system and have some useful resources. Tracking this issue is probably the best way of getting updates, especially to find out when this gets added to the roadmap. While our 2025 roadmap has already been made, making the Android app able to be used by more people is still really high on our priority list!
Pneuma Solutions Achieves SOC 2 Type II Certification, Strengthening Commitment to Enterprise Security and Accessibility pneumasolutions.com/pneuma-sol…
Meta glasses might be useful, but then Meta goes and reminds us all why they're such a terrible, untrustworthy, scummy company. This isn't even the first user privacy violation this year. theregister.com/2025/06/03/met…
I'd still buy them as there is no viable alternative for blind people. I measn… I grew to have this colclusion that #Accessibility is the most important thing to me. I've been telling it to open-source warriors, against-Google, against-Microsoft, against-Meta varriors and so on: if you are abled, you have choice. I'm disabled, I need to work and live my life as fully as I can. So yeah, privacy and all this stuff is good, but to me accessibility is far more vital and important.
@menelion I definitely get it. I'm not judging or condemning anyone for using Meta products. For some it's accessibility, for some it's the only way their friends and family communicate, for some it's for neighborhood watch groups or buying used products. It's sad that nearly all of us are involved with Meta in some way, but it's reality. Plenty of us have no choice.
@menelion I, too, am totally blind. I also agree with you. However, in this case, there are alternatives. I own the ARx Vision glasses. At $599, they are far cheaper than any other glasses for the blind. They also give you the option of whether or not to allow them to use your data for training.
@menelion You can't call someone from them, but they can read text (either quickly/live or after scanning it/document mode), answer questions about it, read bar codes, and tell you what's around you. They can work with the ARx app, as well as Seeing AI, and another app for reading labels whose name I forget at the moment. The page can be found here. Note: I am not affiliated with them. I'm just a user of their product.
The ARx AI Headset: An AI-powered talking camera for the blind and visually impaired. ARx connects to Android, translating the world into clear audio descriptions.
@dandylover1 Yes, no pressure, it's even more valuable for me that you're just a user. Do you know about multilingual capabilities, like can they read French or German or, say, Hebrew or Ukrainian?
I do believe it's multilingual, but I'm not sure. I also don't know which languages it supports. It might also depend on your synthesizer settings. that is, if you have it set to English and try to read French, it will probably read incorrectly.
Help me out here, I want to start a rumor that Asus' new 3000-watt power supply with the 240 volt plug for running four NVidia RTX5090s is a sign that in 2026 you'll only be able to play AAA games in Europe because the American power grid can't handle it.
ASUS Pro WS 3000W supports four RTX 5090 cards at once ASUS Pro WS Platinum power supply is not designed for gamers. Officially, it is meant for “pros”, in fact, that’s exactly what the label on the PSU says.
An extremely expensive, extremely slow motion reenactment of the conclusion of War Games, with open software orgs migrating from one closed-source proprietary service to another over and over again instead of zero-player tic-tac-toe, and me hammering the table saying, come on, learn, god damn it, learn.
The CNCF Slack workspace will be converted from an enterprise plan to a free plan on Friday, June 20, 2025. This will have implications for how Slack works for…
I mean... if the Cloud Native Computing Foundation can't figure out how to self-host a discussion forum for their membership, if the Cloud Native Computing Foundation's membership isn't willing, and able, to learn and support a new shared cloud-based communication system... what the hell are we even doing here? Why does this organization even exist?
I would really appreciate it if you could boost this message. I have created a French version of my Fedi promo video - and a French landing page – that I hope can spread far and wide in the Francophone world:
Cette vidéo explique ce qu'est le fédivers en 4 minutes. N'hésitez pas à partager cette page avec vos amis et vos proches qui ne sont pas ici... J'espère que cette vidéo éveillera leur curiosité 💙🤍❤️
Découvrez un tout nouvel univers des réseaux sociaux, où la vie privée est respectée, les utilisateurs détiennent le pouvoir et les GAFAM n'ont pas leur mot à dire.
doufám, že zítra to bude pěkný. Včera jen ubohých 11km, abych zjistil stav toho kolene. A protože to vypadalo dobře, tak to zítra risknu znovu a uvidíme. Buď to půjde, nebo je i s Alpama konec, protože bych to nedal.
Miroslav Kalousek🇨🇿🇺🇦🇮🇱 𝕏📝💬 @enkocz@twitter.com:
Pravidlo přežití: Nikdy nesmíte přiznat, že se ničíte sami. Vždycky vás ničí zlé a nepřející okolí, nejčastěji novináři.
https://twitter.com/enkocz/status/1935437072564642213
Les précommandes du prochain hors-série Jeux de plateau sont ouvertes ! Pour un numéro 1€ moins cher, et livré dans votre boîte aux lettres : boutique.canardpc.com/common/p…
Blastbay Studios has released two SAPI 5 speech voices. This Web site offers two U.S. English voices, one maile and one female, with each voice costing $19.95. A short audio demo is available for each voice. These voices should work with any program which supports SAPI 5 TTS speech engines, including NVDA, JAWS and Narrator. Please also note that I have not yet purchased these voices so this post should not be taken as an endorsement or recommendation.
I've just downloaded the Libby voice and I am using it now. As with a number of these more human sounding voices, there is a little resistance when navigating at speed even on a high spec PC, and there are some pronunciation issues. But it might be good to have for demos etc as the speech is easy on the ear, particularly for people who are not used to listening to synthetic speech.
I tried out the trial and they're pretty great even with high speeds and very responsive for voices of such high quality. Don't know how he managed to do it but he definitely did.
I have purchased the Richard voice. Screen Reader users should be aware that in order to have it proceed to the PayPal step, you need to press the unlabeled button just passed the "Continue Shopping" button. If you add an item but then press "Continue Shopping" and want to get back to your purchase, press the unlabeled button and it will take you back to your cart with the item in it. This was my experience so wanted to get it out there in case some have trouble purchasing.
Vítejte na mastodonní instanci Mamutovo!
Instance Mamutovo byla vytvořena pro běžné uživatele, kteří touží po klidném prostředí, ve kterém mohou svobodně sdílet své myšlenky.
We expect the #CRTC to release a major Internet competition decision on Friday morning: We think this will be their decision in the various appeals ("review & varies" in CRTC lingo) of their August wholesale competition decision.
Okay, there's a lot going on there, so I'll break it down: 🧵👇🏼 1/8
Well, to make sure we don't have telecom monopolies, the CRTC has long forced the largest carriers to sell wholesale services to smaller competitors like TekSavvy, so we can offer competitive services. That's the basis for Internet competition in Canada. 2/8
From 2015 to 2021, they had ongoing proceedings and appeals over rates and terms. You might remember that because the decision kept going back and forth, and finally ended in 2021 with very inflated rates: crtc.gc.ca/eng/archive/2021/20… 3/8
Attached: 1 image
Update: The Supreme Court dismissed TekSavvy's application for leave to appeal. That means we can't appeal, and the 2021 decision stands, where the CRTC froze the inflated wholesale rates.
Meanwhile, in 2023, the CRTC kicked off a new consultation process about the issues—like they're doing 5-year reviews and each one takes 8 years: mstdn.ca/@andykm/1099882494489…
With new CRTC leadership, this process finally gave us FTTP competition, but there were concerns: mstdn.ca/@andykm/1129566405704… 5/8
The major issue is this: Since the purpose of wholesale competition is to allow small providers like TekSavvy to compete and discipline the market power of telecom giants, should Bell, Rogers, and TELUS be allowed to buy regulated wholesale services? If so, they will use it to further dominate. 7/8
We appealed based on other aspects of the decision too, but they're even more in the weeds (if you can imagine that). The big issue, the one that will shape how internet competition develops, is over BRT's use of wholesale. Stay tuned—the CRTC will publish it here at 9am: crtc.gc.ca/eng/whatsnew.htm 8/8
The #CRTC decision in their own review of last year's wholesale decision is out: The CRTC is not changing the decision. In short, they say the evidence in the appeals did not convince them that they got the decision last year wrong. crtc.gc.ca/eng/archive/2025/20…
Attached: 1 image
We expect the #CRTC to release a major Internet competition decision on Friday morning: We think this will be their decision in the various appeals ("review & varies" in CRTC lingo) of their August wholesale competition decision.
The #CRTC decision in their own review of last year's wholesale decision is out: The CRTC is not changing the decision. In short, they say the evidence in the appeals did not convince them that they got the decision last year wrong. crtc.gc.ca/eng/archive/2025/20…
Attached: 1 image
We expect the #CRTC to release a major Internet competition decision on Friday morning: We think this will be their decision in the various appeals ("review & varies" in CRTC lingo) of their August wholesale competition decision.
The main effect of this decision is to allow Bell and TELUS to misuse this regulatory wholesale framework—a framework that was intended to carve out a sustainable market for small competitors like TekSavvy—to instead dominate those markets and squeeze out independent competitors.
Google Tightens Android Sideloading—At What Cost to Digital Freedom?
Google has begun blocking sideloaded Android apps in Singapore, citing security concerns over apps requesting sensitive permissions like SMS and accessibility services.
Purism makes premium phones, laptops, mini PCs and servers running free software on PureOS. Purism products respect people's privacy and freedom while protecting their security.
Lachtan je v sudetske zoo oblibene zvire. Na papire. Triky uz nejaky patek nedela. Ale kdyz se na vratnici proda dvousta vstupenka, tak je proste potreba vyrazit do prace a neco predvest.
Nacez nadseni divaci cekaji dlouhe minuty s pripravenym telefonem. A lachtan to mezitim zalomil na skluzavce. 🤷
Lawful Access on Steroids: Why Bill C-2’s Big Brother Tactics Combine Expansive Warrantless Disclosure with Unprecedented Secrecy michaelgeist.ca/2025/06/lawful…
Earlier this week, I wrote about how the government’s inclusion of warrantless information demand powers in Bill C-2 may make this the most dangerous lawful access proposal yet, exceeding even the 2010 bill led by Conservative Public Safety Minister …
)
filipa mv
in reply to Xe • • •Xe
in reply to filipa mv • • •Matt Campbell
in reply to Xe • • •Xe
in reply to Matt Campbell • • •