hey @FreakyFwoof for some reason #SendorReadout if I'm connected to an AT&T hotspot, (AT&T here in the US) my network status key doesn't show any information about the provider or the location, etc. IT only shows: Connected: Connected; Status: Up; RX: 39.6 KB/s; TX: 408.1 KB/s; speed: 1.2 Gbps; TSRX: 4.6 KB/s; TSTX: 354.4 KB/s. PM me with what logs you need to try and possibly fix it? OR is this something it's calling to get that info and the provider it's calling for things like provider name, city, state, etc. just doesn't recognize AT&T mobile hotspots, lol.

#AndroidAppRain at apt.izzysoft.de/fdroid/?radd=1… today brings you 19 updated and 3 added apps:

* ShizuCallRecorder: a non-root FOSS call recorder app for Android 11+ 🛡️
* Tuisku: a a simple and lightweight notes app that encrypts your notes 🛡️
* KüA-Plan: access the KüA-Plan (self-organized acitivities schedule) of the CdE e. V. 🛡️

RB coverage: 947 apps (68%) – that is 17 of every 25 apps!

Enjoy your #free #Android #apps with the #IzzyOnDroid repository :awesome:

Sensitive content

Speaking of redundant bootloader features, where is the plug-and-play universal Linux bootloader distro? There should be a Linux distro dedicated for booting. There's LinuxBoot, but its focus is firmware customization, not easy net/multibooting. A "bootloader distro" would ship you a initramfs (or a Unified Kernel Image) as your second-stage bootloader, with its own GRUB-like boot menu and iPXE-like script engine. It would replace nearly all bootloader drivers, and would enable very complicated booting setups, such as booting via Amazon S3 over Wi-Fi.
Please test. A new add-on for BeSTspeech is now out, supporting the multilingual voices. By the way, I spent nearly 2 hours last night trying to figure out the correct sample rate. It's 10800. Not 11025, not anything lower than that. The Russian number was right all along, nobody bothered to notice they were producing lower, deeper-headsize copies of the voice on the 2006 DLLs.
@ChrisDuffley - thank you again for maintaining your fork and making the initial leap to 64-bit. You rock. Pleas review the PR with or without AI, it don't matter as long as you get what it does in your way. github.com/ChrisDuffley/b32tts…
you may find the copy here: eurpod.com/synths/bestspeech_2…
This entry was edited (today, 8:28 PM)

reshared this

in reply to Tamas G
thanks for doing this. It's soooo weird hearing this thing speak Polish. It's good you kept the old version around too because these new libraries seem to be very unstable and keep needing to restart the host process. Tracks with when I tried using the demo speak window though which also didn't like longer strings.
in reply to Pitermach
@pitermach wow this is already really good feedback - I wonder if we'll have to do some type of chunk splitting for the V2 DLLs. Keep stress-testing it, especially with say-all and large strings to see, because those are the things worth guarding against before a PR is merged.
in reply to Tamas G
Perhaps yeah. Found a more serious bug which may be with the Polish language, maybe others too but I haven't checked. If you have use spelling functionality checked and read some letters like C or D, the synth stops speaking, though NVDA and braille keeps working. If I try switching to another synth in this state NVDA locks up.
in reply to Pitermach
oh now that's interesting. Just tried here and typing either letter didn't crash Polish for me. Wow how interesting.
Still putting it on the radar though. If you can send over a log with debugging enabled of just that (privately works too) I'll investigate that.
This entry was edited (today, 8:10 PM)
in reply to Pitermach
@pitermach I stress-tested 255 through 8000 characters: classic handles all of it, but both V2 DLLs crash the helper dead on anything over 252 characters — a fixed 256-byte text buffer (Japanese's is roomier at ~580). NVDA sends paragraph-sized utterances constantly, so this was the "keeps needing to restart the host process" instability. The wrapper now chunks V2 text below the limit at sentence and word boundaries. One subtlety mattered: I verified tilde commands reset on every Say_TTS call, so the leading command run is parsed off and re-applied to every chunk (proven working across a 20-sentence utterance), hard cuts refuse to split inside a mid-text command, and chunking checks the cancel flag between chunks. The full sweep now passes with zero failures. terminate() joined the background thread before killing the helper; a hung synthesis left that thread blocked reading helper stdout, so the join deadlocked NVDA forever. The helper now dies first, the read gets EOF, the join returns. Even if some future poison input hangs an engine, NVDA stays responsive. So. Here you go:
eurpod.com/synths/bestspeech_2…
Buenas, perdón por el retardo:
github.com/JosePerezHuanca/NVI…
El complemento es compatible con NVDA de 64 bits pero se mantiene la herramienta externa por retrocompatibilidad.
Casi lo descarto pero creo que fue mejor así.
Traducción al inglés así que ya estoy listo para publicarlo en la comunidad internacional si es que no encuentro errores a último momento jajaj.
in reply to yares
Como el complemento cumple con su propósito totalmente tengo algo que me da vueltas.
Si el año que viene no hay nada que corregir, a partir de ahora toca actualizar buildVars y suvir tag una vez por año.
Podría no hacerlo, dejarlo así y el complemento funcionaría si alguien fuerza la compatibilidad, pero son esas cosas que me incomodan de la API de NVDA porque creo que generan desconfianza para el usuario promedio.
Publicar una release solo para cambiar dos numeritos me parece insignificante.
This entry was edited (today, 7:58 PM)
in reply to Friedri Schmerz
Jugendlicher Idealismus- der eine Vorstand im Ort, in der Region, der wirklich was auf dem Kasten hatte. Frau Süssmuth und manchmal Herr Günther. Menschen, deren Position man respektieren und mit Respekt begegnen kann. Die, wie man selbst an der Demokratie interessiert sind. Und dann kommt Herr Spahn, Herr Merz oder Frau Reiche und der Idealismus, die Energie ist flöten 😚

Qué sinvergüenzas que son.

El Gobierno de Ayuso rectifica y dice que venderá el ático de lujo para financiar la recuperación tras los incendios un día después de conocerse su compra eldiario.es/1_ccbd5a

Planifica Madrid, se llama la empresa, amigos. Ahí tiene que haber tomate y pastuqui y paquípallá.

La empresa pública que compró el ático para "uso temporal" de Ayuso lleva un año tratando de vender un inmueble mayor en Gran Vía eldiario.es/1_ccbd8b

"Total, las iban a asesinar igual..."

El Gobierno de Ayuso ha destinado 849 euros para prevenir y sensibilizar contra la violencia de género en lo que llevamos de 2026: el presupuesto anual sufre un recorte del 50% cadenaser.com/cmadrid/2026/07/…

Esto no tiene fin, ¿eh...? Y ahí van los beneficios del trabajo bien hecho y sudado.

El novio de Ayuso pone a la venta el piso en el que vive con la presidenta eldiario.es/1_ccc128

After BookAmp comes BookBar — the same features, this time for foobar2000: tempo adjustment, pitch adjustment, both at once, a limiter, and audio smoothing with filters.
github.com/michaldziwisz/bookb…
wow so I must be the only one for whom the iOS DoubleTalk with the ROM loaded from my iPhone into it, won't speak. Seeing a lot of successes. For me, restarted phone, toggled VoiceOver on and off, nothin'. And yet "preview voice" speaks fine. Maybe it's only since I'm on iOS 27, ha.
in reply to Chad Blanco
yeah, I haven't looked into whether Mame has a library-only version of their emulation layer, like Unicorn with Qemu- but Qemu already does do a lot of CPU chip types, so it could be very useful elsewhere in the toolbox later on, potentially. Just not for Apple II and that family. Whoever figured out how to get SmoothTalker working I think had to know some good background on speech synths, unless the AI really came up with that idea of memory-snapshotting the .com at runtime without DOS underneath so we could pipe audio into it freely. Somehow I'm thinking not, but these days it wouldn't surprise me if it did 80% of the lifting on that solution.
This entry was edited (today, 8:05 PM)
LinkedIn adds a button to report AI-generated ‘slop’ techcrunch.com/2026/07/30/link…

this isn't to report AI slop

this is to help tune their model.


LinkedIn adds a button to report AI-generated ‘slop’ techcrunch.com/2026/07/30/link…

New slop just plopped like a big full belly, and she jiggles real fine like Santa's bowl of jelly. ... Or something.

Anyway, for many years, Linux has been stuck with two TTS engines, ESpeak and IBM TTS, and we recently got DecTalk. Recent as of like 2 or 3 years ago, which is basically yesterday in Linux #accessibility terms. So, this morning, I had #ai take some TTS NVDA addons and port the TTS engines to Linux, Speech Dispatcher, to be used in Orca. And they all work. Amiga Narrator sounds weird, but the rest are perfect. And this was 2 hours of casual prompting with GPT 5.6 Sol at low effort. Those who hate AI, I don't have anything left to say to show how much it has helped the blind community. Also if you come at me all mad and stuff, I'll eat your breakfast. Right in front of you.

- Repository: devinprater/retro-tts-pack (github.com/devinprater/retro-t…)
- First release: Retro TTS Pack 0.1.0 (github.com/devinprater/retro-t…)
- Linux x86_64 tarball (github.com/devinprater/retro-t…)

#blind #tts #linux #tech

Я всё ещё не могу вспомнить и нормально ставить правую руку для игры различных арпеджио и прочего, поэтому решил попробовать начать учить песню Сансары "Падаем", в которой используются аккорды и бой. У меня получилось поставить баре в фа-мажорном и си-бемоль мажорном аккордах. Пытаюсь понять, как играть бой. С этим у меня всегда были сложности. Интересно то, что после игры 3 дня подряд и паузы в 2 дня у меня перестали болеть пальцы на левой руке, видимо мазоли уже по-тихоньку начинают образовываться так, чтобы не болели пальцы, это мы проходили. Пока немного болит левая кисть после попыток поставить баре много раз, буду думать и смотреть, какие лишние движения делаю, ну и правая рука также некомфортно себя чувствует при игре арпеджио из-за непонимания ошибок в постановке.
in reply to Alexey Skobkin
@skobkin У меня почему-то не хватало не то сил, не то сноровки нормально прижать струны, и звук получался… ну такой, блямс, глухой очень, яркости не хватало. А сейчас у меня гитары давно уже не было в руках.
Anyone else having trouble lately with Google Messages for Web? It seems like it is connecting to my phone and even reading new messages, but the actual web page shows the can't connect error and I can't arrow to any conversations. Some are saying the latest Messages betas are causing this so may try uninstalling them.
This entry was edited (today, 7:36 PM)
The media in this post is not displayed to visitors. To view it, please go to the original post.

Breaking! 🚨 Bueno, tenemos un par de anuncios que hacer, el primero es que el tan anunciado servidor de respaldo cifrado ya quedó operativo ayer, junto con el servidor OpenBSD de la comunidad Tilde UnderNET. El consumo eléctrico aumentó sensiblemente pero esperamos poder cubrir eso con las donaciones mensuales que recibimos. El almacenamiento aumentó en 3.5TB redundantes y ahora tenemos un total de 26.5TB, 34-cores y 104GB de RAM. El consumo calculado ahora es de unos 450kWh un equivalente a unos $2900 mensuales de UTE IVA incluído aprox. Ahora si, no vamos a agregar más servidores. El siguiente anuncio lo explica @j3j5 👇 #mejoras #uruguay #autogestión #mantenimiento
This entry was edited (today, 5:01 PM)
in reply to Santiago 🔭🪐

Como ya es conocido para los que siguen un poco nuestras andanzas en Undernet, tuvimos unos cuantos problemas con scrappers y otros actores maliciosos. Gracias al esfuerzo de @santiago, ahora hay toda una línea de defensa que bloquea direcciones IP dependiendo de una serie de reglas (intentos de scrapping, intentos de exploits, login por fuerza bruta...).

Esta es una lista de "atacantes" que, al intentar dirigir sus ataques contra Undernet quedan bloqueados en el firewall, no es una lista de fuente pública (que también usamos). Pueden consultarla aquí:

undernet.uy/blocklist/

Ahora, algunas veces los sistemas de defensa se pasan y pueden bloquear a alguna persona que no hizo nada malo (excepto usar mucho Undernet, ejem @gabriel, ejem). Pues gracias a que la lista es pública, armé una paginita que te permite comprobar si estás incluido en los bloqueos:

undernet.j3j5.uy

Por motivos obvios (si estás bloqueado, no podrías verla), esta página no vive en el datacenter de Undernet, es mi pequeña contribución al ecosistema Undernet.

in reply to Julio J.
Se queda pendiente hacer algo más con los datos, si me dan las ganas, me gustaría armar algún tipo de bot o mini-dashboard para ver el "nivel de amenaza" contra Undernet (que con un poco de brocha gorda se podría extrapolar a un nivel de amenaza país). Ver qué países son los más bloqueados, cuántos reincidentes hay...no sé, hay muchas opciones. Como siempre, escucho sugerencias y comentarios :)
The media in this post is not displayed to visitors. To view it, please go to the original post.

‼️¿Quieres formar parte de Terebi?

Escríbenos a info@terebimagazine.es

➡️ Mándanos algo que hayas escrito
➡️ O tu portfolio si eres ilustradore

⏺️ Sobre el texto o portfolio:
- NO hace falta que se haya publicado en ningún sitio
ni experiencia previa en ningún medio

⏺️ Sobre el proyecto:
- NO es remunerado, depende de Patreon
- Funcionamos con un calendario anual, se escribe un mes y se descansan dos

📝 Abierto el plazo hasta el 17 de agosto. 🎨

🎮¡Si tienes cualquier pregunta, escríbenos!🎮

The media in this post is not displayed to visitors. To view it, please go to the original post.

¿Recordais este meme que lleva ya un tiempecito rulando? Pues por lo visto el Sam Altman acaba de decir lo mismo pero calcao, con las empresas de IA... 😅
The media in this post is not displayed to visitors. To view it, please go to the original post.

OpenAI makes two GPT-5.6 models cheaper, expanding usage in ChatGPT 9to5mac.com/2026/07/30/openai-…

reshared this

The use of AI in UK asylum decision-making is a disaster.

AI hallucinated a document that led to an asylum application being denied.

Applicants aren't being informed that opaque AI tools are being used in their case, nor being given the opportunity to correct errors churned out by these opaque tools.

theguardian.com/uk-news/2026/j…

#ai #asylum #immigration #migrants #homeoffice #ukpolitics #ukpol

cruelty is the point of #fascism. #ai is the automation of cruelty.


The use of AI in UK asylum decision-making is a disaster.

AI hallucinated a document that led to an asylum application being denied.

Applicants aren't being informed that opaque AI tools are being used in their case, nor being given the opportunity to correct errors churned out by these opaque tools.

theguardian.com/uk-news/2026/j…

#ai #asylum #immigration #migrants #homeoffice #ukpolitics #ukpol


Before I start trying to figure this out myself: Is anybody familiar enough with the internals of the DoubleTalk emulation, or the DoubleTalk chip/card itself, to know what it would take to make pronunciation dictionaries that would be reasonably portable between the NVDA add-on, iOS/Mac OS apps, etc? I honestly find myself using it more than I realistically expected, but it does have its limitations and I run into weird pronunciations all the time that I wouldn't mind fixing. Adding every single one to VO/NVDA dictionary is kind of annoying especially across Mac and iOS, and doesn't give me direct phoneme access or, particularly on Apple systems, much of a way to control when to apply the rule (e.g. through regex). A community dictionary, plus the ability to import custom user dictionaries, would of course be the best option. Boosts appreciated!
This entry was edited (today, 5:50 PM)

reshared this

in reply to Guillem Leon
I realize that this is an issue for every synthesizer, but I figure this is one worth tackling given the nature of the emulation and how much control we all have over it. And also because DoubleTalk is particularly finicky with its letter-to-sound rules.
in reply to Andre Louis
I am one of those people that if I am outside, and someone drives by with there stereo up, but it sounds like shit, I yell don't blair it if you don't got it! lol! I actually helped a friend of mine install the system in his mustang that he used to compete in a bass compitition. I couldn't help with wiring to the amp, and to the car, but I helped him with how it sounded, and helped him with the raddeling trunk. anytime he made changes to it, meaning equipment changes, he asked me how things sounded, and let me adjust the E.Q for him. those were the good old days.

all I've ever wanted for books, movies, and videogames is to be sent a physical copy when I buy a digital one.

I buy the digital one so I don't have to drive or wait to access it. But I still want the physical copy. Why can't they give me the digital copy but still ship me the physical for the same price? How is that considerably more expensive for them to do? The bandwidth is damn near free for them

Planet GNOME now shares blogs to GNOME Discourse! However, owners of blogs need to first opt into it; we treat consent seriously.

You can read them and engage with them here 👉️ discourse.gnome.org/c/communit…

#GNOME #Discourse #RSS

The media in this post is not displayed to visitors. To view it, please go to the original post.

Oh my. A trusted source who enabled Google's new AI feature for Gmail just received this nudge from the service today, which encouraged him as part of his suggested to-dos list to fall for a cryptocurrency scam waiting in his inbox.

IDK why, but when I saw this I was reminded of that Seinfeld episode where Kramer gets a new phone number but it's the old Moviephone number, and so he starts answering the phone and reading the listings. "Why don't you just tell me what scams to click?"

youtube.com/watch?v=XagGEi_n_o…

in reply to BrianKrebs
If your banker, broker or other trusted adviser lured you into fraud or robbery, you can bet they would be held responsible for their actions. Just because the trusted confidant has chosen to replace measured analytical thought with a pseudo random statistical algorithm shouldn't invalidate their obligation to behave responsibly.