Friendly reminder of why #Google is just plain evil: they've taking down #IzzyOnDroid (a third party Open Source Android "app store" like #FDroid) *twice* in the past 30 days over false positive malware scanning.
We are receiving reports that Google flags our repo as "having dangerous apps" or being a "dangerous site" – texts being very vague, no proof given (nor did they inform us). They also link to a page they call "Transparency Report" – which is of the same vagueness, but definitely not transparent (transparencyreport.google.com/…)
We're not aware of any such dangerous content. All apps on our repo are properly scanned, see izzyondroid.org/about/security…
it's things like this, which hinder the adoption of open source software. I today had an unproductive discussion online with someone who sells why Apple products are the best out there and I annoyed that guy by saying that Linux on a laptop could perform better, but the thing it's always an uphill battle.
Hardware companies not cooperating, laws, no funding, no market, etc.
@phantasus Yeah. Linux is great! But how is anyone ever going to find out when you can barely ever buy it preinstalled (at least not in any "normal" store)?
And then with all the extra fighting we have to do with smaller teams and less budget to not have our existence be censored by the big players, we are even expected to meet higher standards than them (nobody calls Google Play "scary" despite all the malware cases they've had while we had none...)
Right now, the frontend wil compile on all platforms, including Linux. SpeechPlayer itself though relies on windows.h so it does not compile, so that would need to be made cross-platform, for those asking. I purposefully tried to keep Frontend very platform-agnostic to avoid headaches later.
I do want to add Gemination though, kind of itching to for Finnish, Hungarian, Japanese, these languages use it, and without rules that help them, they just won't sound good on their own unless you do lots of YAML hacks. Also that word sounds cool. Gemination and geminates.
It's really nice now. Like, all I need to do for a new setting is: 1. Create passes/liquid_dynamics.cpp, or whatever rule I want. 2. Add my settings to LanguagePack 3. Register it in pass_pipeline.cpp 4. Done The core engine stays clean. Each rule lives in its own file where you can tune it, break it, fix it, without touching anything else. Honestly this would be an old 1990s TTS engineer's dream and that's a good feeling.
“Can we just stop equating success with short-term economic growth? What if we used usefulness and longevity? What if we gave more value to the fundamental technological infrastructure instead of the shiny new marketing gimmick used to empty naive wallets? Well, I guess that if we changed how we measure success, Europe would be incredibly successful.”
Y'all I want to get in the hot tub and not code for the day I hope that's acceptable. We added a lot of stuff today, now I hope languages can take advantage of these new settings. Only by adding good structured rules can we make this thing sound better. Now that the frontend is moduler. So I'm going to update packs from here on out probably, mainly English ones, make words like "good" and "book" sound less British, stuff like that, but I really want to lock the frontend down a bit with some more modularization like this and then worry about extending it as languages need it. It's now as easy as adding a new CPP in /frontend/passes for any extr things, that makes me very happy. I'm not going to keep bloating ipa_engine.cpp with more rules.
Cate Blanchett as Steve Jobs Natasha Lyonne as Steve Wozniak Maya Hawke as Susan Kare Kathryn Hahn as Phil Schiller David Tennant as Craig Federighi Shohreh Aghdashloo as Johny Srouji Emma Thompson as Johny Ive
and Tilda Swinton as Apple Legal. All of them. Every single one played by Tilda.
yep, he never pointed it at anyone, never even touched it, they just used him carrying as an excuse to execute someone being a slight nuisance in their presence
New NVSpeech Player update! Huge one. But it's V13 so maybe it'll be unlucky. Who knows! I did a major engine change, but languages are kept the same for now. You'll see checkboxes for coarticulation, microprosody, and phrase-final lengthening. Here's what they do — and why the difference is subtle but real. Coarticulation: Your tongue doesn't teleport between sounds, it glides. These settings make consonants "lean toward" neighboring vowels. The "velar pinch" makes /k/ sound different before "ee" vs "oo" — try "key" vs "coo" and notice how the /k/ changes. That's real, and now the synth does it. Microprosody: After voiceless sounds like /p/ or /s/, the next vowel's pitch starts slightly higher. After voiced /b/ or /z/, slightly lower. Maybe 10-15 Hz — tiny, but expected by your brain. Phrase-final lengthening (off by default, useful for some languages): We slow down at sentence endings. This stretches the final vowel so speech feels punctuated, not abruptly stopped. Rate-dependent reduction: At high speeds, unstressed schwas get shortened, keeping fast speech crisp. Anticipatory nasalization (off by default): Vowels pick up nasal resonance before /m/ or /n/. Optional since languages vary. Will you hear a dramatic difference? Honestly, each one is subtle — like color-grading a photo. You might not spot what changed, but it feels less synthetic. I encourage language speakers to mess with them for their language and report which numbers work. eurpod.com/synths/nvSpeechPlay…
Also liking Robert with the 16 khz sample rate. Reminds me of the Toshiba TTS that used to come with old Toshiba laptops, if you've ever heard that one.
@miki oh, both from people's Feedback being critical but genuine on how things sound in their language and then 3 different AI systems (Claude, Gemini, GPT) to pass code through and ask research-level questions about sounds and phonetics. I feel like I've absorbed months of linguistic knowledge in a matter of about 2-3 weeks, LOL. How strange to learn this way, way more hands-on than first reading all the papers and you learn a lot by ear too and just gain a deeper appreciation for human speech and language. Honestly that's probably why I'm sticking to it in the first place for now.
@Alan nah, but the 2026 one is still compatible! I've kept it around mainly to keep a version locked down, but truthfully maybe it could confuse people into thinking they can't install the 2026 one on NVDA 2025.X and below, so I'm considering removing it, but then I'd break lots of old links too, ah the quandery
I have another question. Is there a parameter I can adjust to change the gap between consonants? I find it hard to understand because they are all squished together.
@Alan so stopClosureMode — controls when closure gaps are inserted. If you want maximum separation, try "always". If you want “mostly, but not everywhere,” keep "vowel-and-cluster" and tune the gap times. stopClosureClusterGapMs — the size of the gap used in clusters.
Tomorrow, provided both hosts have power, the Digital Domain with @nick and Bob is also known as the 'Bring on the Snow' show! How was Snowpocalypse 2026 for you?
Also, predicting the grammies and listener submissions that fall into the category of WTF?
Fun fact: about 2/3 of the apps in our repository are confirmed as #ReproducibleBuilds – and (rough guess) about 90% of those are hosted at Github.
Now: Github being marked "malicious" by Google "Safebrowsing" – when? That's where all the malicious code lives then, no? And where all those (90%) "malicious" APKs are attached to releases (which we scrutinize with a bunch of extra scans; you can compare the file hashes to confirm) 🤔
And for the first time two totally different clients are sharing Stories, in real time, without any adaptation in between ✨ ..and everything is stored on your XMPP account! 🔒 Checkout the Stories integration in @monocles 🤩
Attached: 2 images
Today, the next release of monocles chat is finally ready and published, and will be available in stores in the next few days. It is a huge update with three fundamental new features:
Stories, Feeds, and Phone log.
I have started using @Tutanota and I am pretty satisfied with it except for a big issue, which might be a big barrier for many people. You cannot import email unless you buy the Legendary subscription. I bought Revolutionary, and what a surprise! I have to say now, if you go to Pricing in the website, there is no mention to that. In fact, when you try to import (only possible in the desktop app) you get an invitation to change to Legend, with the same information, so you cannot say what is happening. If you go to the help center (Support), there is no information in the Import sections about the account level. In fact, I'm currently unable to find the page where Tuta explains that by the moment only Legend accounts can import email. I totally understand that migrating accounts can be an important overhead for Tuta now. I also see email importing like a critical feature. It should be included in all paid subscriptions. Maybe limiting the migration size or rate, but it should be present. And, in any case, it should be clearly stated in all the help pages. Now I have an odd and not very positive feeling about Tuta, like this company was hidding very relevant information when offering their products. I'd like to import my old emails, but I really do not need the Legendary subscription. Indeed, I might accept paying an entry price enabling me to migrate my old account. I have to say, you need to take care of the help system too. For instance, in the "forward email" section, you show links with further information about activating email forwarding from different providers, but these pages are only comparisons between Tuta and the other providers, there is no explanation about email forwarding. #Tuta #EmailImport #drawbacks
thank you for your feedback! I sound more critical than I feel, but I really think that the information should be clearer in the help pages. I can wait, I want to use Tuta for personal emails and I can manage by forwarding messages and not closing my old accounts by now. In any case, thank you and keep the good work!
Hey @Tutanota, I have been using your Tuta mail for some time now. I love the concept of encrypted emails and have since gotten many people to switch to your services. The only problem I have is with my partner, with whom I share a calendar. When a new event is created, the other person is not notified. Is there already a solution for this, or are there plans to implement something like this? Thanks a lot in advance and btw. I live your presence here in Mastodon ❤️
With a shared calendar, there are no notifications about new event. If you want that, you can invite them to a calendar event (without sharing the entire calendar), then they'll get an email through which they need to accept the event.
Hello @fastfinge and others, I added two new tools to /tools in the SpeechPlayer Repository. Let me know if these can help you: formant_trajectory.py — The Visualizer This is an accurate Python port of frame.cpp's FrameManager. python formant_trajectory.py --packs /path/to/packs --text "hello world" --voice en-gb --out trajectory.png --wav audio.wav python formant_trajectory.py --packs /path/to/packs --ipa "həˈləʊ" --out trajectory.png --show # Vowel space plot python formant_trajectory.py --packs /path/to/packs --text "beat bit bet bat bot" --vowel-space vowels.png frame_inspector.py — The Debugger For when you need to understand why a transition sounds wrong: # Dump a single phoneme's frame parameters python frame_inspector.py --packs /path/to/packs dump a
# Compare two phonemes side-by-side python frame_inspector.py --packs /path/to/packs compare a i --fade 10
# Trace the exact interpolation sample-by-sample python frame_inspector.py --packs /path/to/packs trace a i --duration 100 --fade 10 --out trace.png
# Find the most "violent" transitions in your phoneme set python frame_inspector.py --packs /path/to/packs grid --phonemes a e i o u p t k s --metric f2_rate
# Estimate consonant locus (for coarticulation tuning) python frame_inspector.py --packs /path/to/packs locus p --vowels a e i o u
I last did lap swimming in early 2020. I stopped when the pandemic started, and I didn't go back... Until today. I've wanted to for a while, and today, things worked out such that I was able to get a ride there and back. You guys, I think I went too hard. I aimed for 1000 yards, but wound up feeling good enough to go to 2000. Now everything from my mouth down is sore. Yes, my mouth. I forgot about all the mouth work involved in blocking and expelling water.
@miki Nope. I adjust my strokes a bit so my hand hits the wall first. So far, I haven't hit my head. I also change the angles of my hands so my fingers don't hit first. Counting strokes is a good idea. I know that after about 20 strokes, I should start bracing for the wall. No flip turns, just find the wall, reposition, and push off again.
One of the notable aspects of #Mastodon is that even when discussing highly controversial topics -- with rare exceptions -- it's possible to get up in the morning, check your replies, and see a completely civil discussion. Try that on most social media. Good luck.
The US has become fascist, and we Germans can say that... we have seen such surveillance methods being abused twice, in the past, with pen, paper and wiretapping. #Stasi #Gestapo #Thiel.
It's clear I'm going to have to modularize the frontend. As such, it will now do various passes for these rules and prosody changes. But I'm going to try and keep things sounding as they are unless new settings are turned on for people. Legacy Pitch will still work too don't worry.
hmm. My approach will be to increase the caNP value for the vowel when anticipatory nasalization is enabled and if a nasal consonant follows within the same word or syllable. This is interesting research.
reminds me of the time we had an AC failure in a part of the datacenter in dead of winter so we opened the doors ... and froze the hard drives in the first row of servers
🎙️ Fresh podcasts this week on Castopod Index! Software quality, digital security, open-source business tools, veteran stories, low-tech viticulture, AI in construction & more! 🌍 Every podcast is on the Fediverse: follow, comment and share directly from Mastodon! 🚀 index.castopod.org/new.html
Would it be too much to ask if you could implement an automatic update system like WordPress has, which notifies you every time you update the CMS? In other words, we'd like a notification in our dashboard that a new version is available, and we could simply update with a single click, without having to go through tedious backup procedures, deletions, file downloads, needing an FTP client, etc. (especially since these are often minor updates). Thank you.
would it be possible for podcasts to be built on #ActivityPub in addition to RSS, and for podcatchers to process it? then each episode would be its own AP item and we could boost them directly, not just link to them. I feel like that would open up a lot of other possibilities too. #podcasting2
Finally found a thing mostly resembling what I imagine a web zine to be.
I want to create something similar.
The criteria:
⁃ pulished in issues ⁃ sequential editorial design resembling printed media ⁃ simple and robust web technology only, should be readable on 10+ years old devices ⁃ niche and independent af ⁃ delightful ⁃ downloadable/replicable ⁃ privacy respecting
It's also clear I need improvements for Geminate consonants which Hungarian uses, this is why Hungarian and some languages still sound off. these will be new rules. Sue Hertz made Eloquence with lots of rules. I must consider how her work was and read papers. What if we added: geminateClosureScale: 1.8 geminateReleaseScale: 0.9 # release stays normal
Phantasus 🏳️🌈
in reply to Sylvia • • •it's things like this, which hinder the adoption of open source software. I today had an unproductive discussion online with someone who sells why Apple products are the best out there and I annoyed that guy by saying that Linux on a laptop could perform better, but the thing it's always an uphill battle.
Hardware companies not cooperating, laws, no funding, no market, etc.
Sylvia
in reply to Phantasus 🏳️🌈 • • •@phantasus Yeah. Linux is great! But how is anyone ever going to find out when you can barely ever buy it preinstalled (at least not in any "normal" store)?
And then with all the extra fighting we have to do with smaller teams and less budget to not have our existence be censored by the big players, we are even expected to meet higher standards than them (nobody calls Google Play "scary" despite all the malware cases they've had while we had none...)
FOSS is never given a fair chance :(