Don't get me wrong. I like AI music, I think it can do some really cool things, but but I don't want it on my spotify. I'd rather have it on suno or other platforms that specialize in it. Use AI music like a scratch pad. But that's just my opinion.
That was a great #FOSDEM again, met so many old friends and new people, saw interesting talks and had good technical discussions and some new ideas for things to try out now.
Some people I didn't run across, but the next event will come soon. :)
See you next year hopefully.
I'll be in Brussels for the #buildroot developer meetup still for the next three days.
@Scott I got it sorted (for now) I'll bug you later though. It is just silly stuff about setting up the git repo so I can move things from a dev branch to a main/release branch and preserve the commit history.
The neighbors are clearing our driveway of snow. There are four people in the house, and we're all in various states of not able to do it for different reasons. The little harf harf dog is not happy.
@ricky_enger well, the problem is we got enough snow that it pretty much completely buried him when he went outside, and he didn't really like that much.
@ricky_enger The dog my parents had before Watson (Dexter, the 25 pound Schnoodle) hated snow with a passion, until one day, he just didn't anymore. Don't know what changed for him. Too bad he died not long after.
Ireland has dismantled its first commercial wind farm to make way for a step-change in scale. At Bellacorick Wind Farm, 21 early-generation turbines have been removed so Oweninny Wind Farm can expand. Each new turbine now generates more power than the entire 1992 Bellacorick site, illustrating how three decades of innovation have radically compressed land use while multiplying output. RTE rte.ie/news/regional/2026/0116ā¦ #ShareGoodNewsToo
NVSpeech Player with phoneme editor and NVDA Addon, Speech Dispatcher module version 2.5 is out! The focus this time: more control over voice quality.
New in DSP Version 5: The engine now has 12 voicing tone parameters for shaping the overall voice character, plus 5 new per-frame parameters for things like creaky voice, breathiness, jitter, shimmer, and glottal sharpness. That last one is key for getting that crisp Eloquence-like attack some users prefer. For NVDA users, this means 10 real-time sliders in voice settings instead of 4. Dial in your preferred sound without editing config files. The Windows phoneme editor has a new Voice Quality section with full access to all the new parameters. Useful for anyone tuning phonemes or building voice profiles. Linux users get an updated renderer with command-line flags for every new parameter. Speech Dispatcher configs are included. Voice profiles from phonemes.yaml now work across platforms. While no change yet for Danish support, the foundation is now here for it in the future. github.com/tgeczy/NVSpeechPlayā¦ github.com/tgeczy/NVSpeechPlayā¦ github.com/tgeczy/NVSpeechPlayā¦
This is a five-year-old Mac. There's just one thing that can consistently make it feel sluggish (beyond the obvious ones like running a big build with all cores blazing). What is it? It's obviously Win11 on Parallels.
Sunday. made the daughter happy by buying another accessible card game for the collection. It's a Mille Bornes knockoff but was very reasonably priced. 64ouncebraille.com/products/10ā¦
[This game contains braille already on the cards and no additional setup is needed] In 1000 Yards the kids compete to be the first one to go 1000 yards.Ā Kids use skates, skateboards, bikes, scooters and more to make it.
ā¢ You wake up late, but still feel tired ā¢ Every day feels the same ā¢ You apply for jobs daily, no replies ā¢ Money feels tight all the time ā¢ You feel guilty even while resting ā¢ Confidence slowly goes down ā¢ You start doubting yourself ā¢ Skills are there, opportunity is not ā¢ Nights are full of overthinking ā¢ You feel stuck and helpless ā¢ You are trying, but no one sees it
Being unemployed is not laziness. Itās mental pressure no one understands š.
On my way back from my parents' to Vienna and I was driving through the Czech Republic. In tribute to that fact, have my first NVDA Composer thing - Jozin z Bazin by Ivan Mladek, because we all need some more songs about swamp monsters called Joseph, or something... Could be better but given that it took most of the 4-hour ride to put it together, I don't think it's bad. Haven't found the special Nokia mode with the wav exports yet which could have saved it. Shall we call this #Composituary?
Just bought a Lenovo IdeaPad off FB Marketplace. It seemed like a good idea, since, at some point, I need to work on infrastructure to allow Orca to use touch gestures, unless someone else gets to it before me. Therefore, I need a laptop with a touch display. So now I have another laptop to set up.
@tyrylu I feel somewhat busy right now, but I'm wondering what you have in mind. I'd definitely want to be aware of the relevant use cases. I just sent you an email.
A wave of manic energy in December had me put together a long deck called "HarfBuzz at 20! " , celebrating 20 years of HarfBuzz. š
I designed the deck to be presented at the #WebEnginesHackfest later this year. Then reality hit that I cannot present this deck in any sane amount of time.
Inspired by all the great presentations coming out of #FOSDEM today, I decided that instead of tossing the deck out, I just put it out here to be read by the curious. I will present a highly condensed version at the hackfest in June.
@Friendica Support Wenn ich einen ƶffentlichen Beitrag zusƤtzlich zu meinen TL-Kontakten auch einem Email-Kontakt schicke, erhƤlt dieser Email-Kontakt auch Email-Benachrichtungen Ć¼ber alle Kommentare?
Zu diesem Thema: Kann ich einen Email-Kontakt dauerhaft im Editor eingeben? Falls ja: wie?
Danke schƶn fĆ¼r Eure Hilfe . . . Gerhard / oldie
"But Wikipedia cannot survive by refusing to change. Ostromās Nobel prize-winning research reminds us that the communities that govern shared resources often grow conservative over time.
"For anyone who cares about the future of reliable information online, Wikipediaās 25th anniversary is not just a celebration. It is an urgent warning about what happens when the institutions we depend on cannot adapt to the people they are meant to serve."
This week I learned, from a blind acquaintance here, that when we write in ALL CAPS in alt text, their screenreader reads it all out, *one letter at a time*.
Totally defeating the point of alt text making things accessible, right? Needless to say, I've stopped.
@menelion huh. What screen reader does that? Is this an iOS/Android thing? I'm a screen reader user myself and I don't think I've seen this happen all that often
@zersiax I'm quiet here, because I use JAWS and I've seen it before maybe, but my hunch is that it depends on the TTS. I boosted though because I feel all caps is like yelling, also it adds burden for Braille display users.
@menelion I do use all caps to emphasize words in a sentence, as in "Why DO chickens cross the road?" but that's about it. All-capsing a whole entire post ... ehh ... I feel it looks like someone feels the need to scream but then, I don't mean to tone police either, just not my thing
Some really great reads for January. I had no idea Dennis E. Taylor's standalones were so good! Not to mention a new Peter Cawdron and a couple of interesting fanfics.
@BorrisInABox Yeah I really enjoyed these too. His stuff is just good fun. Bobiverse starting to go a little stale maybe, so glad heās exploring other ideas. Pity that, unless Iām mistaken, he seems to be exclusive to Audible though.
if (frameExSharpness > 0.0) { baseSharpness *= frameExSharpness; baseSharpness = clampDouble(baseSharpness, 1.0, 15.0); That way you can't break the sharpness slider. I think this can work well. AI was wanting to make negative sharpness. Oh fuck no.
You can tell if someone is a computering supergenius if their solution to a difficult problem looks like nothing.
Lisp is six functions. Forth is 200 bytes. Unix is just tiny programs and text files. The original web is just a hacked SMTP server sending SGML files. And yet, it does *that*.
The huge, complex stuff--Windows, Java, the modern web--is all the work of mediocre thinkers with big budgets and too little time.
I tried to build this small python tool on FreeBSD that uses the new uv thing but I couldn't figure out how to work uv
They really need to work on their UX
All I know is that it kept complaining that there wasn't any Python 3.13 but I couldn't get it to download or install one (into a virtual env I assume). Now I am aware that they probably don't have precompiled versions for FreeBSD, but it doesn't even seem to have an option to download the source and compile it by itself. So if that's the case, this tool is pretty worthless if you want to use it with a custom python version or on an unsupported platform?
Is that really how this thing works? Bizarro
It was easier to just hack it to work with my own venv and regular pip, build my own requirements.txt etc
Provided to YouTube by Universal Music GroupSweetness Ā· Jimmy Eat WorldBleed Americanā 2001 Geffen RecordsReleased on: 2001-07-17Producer: Mark TrombinoProdu...
@Scott I do still hear it (or spin it) from time to time, but I can't really account for it today. I did watch "Spider-Man: No Way Home' last night; maybe seeing Toby Maguire's Spidey again evoked the era?
I think we're going to add voice sharpness as a frame parameter for the new frame. this would allow a lot more things to happen. Emotional emphasis, Creaky ā sharp transition (stĆød), Consonants sharper than vowels. If we make it a frame param, it can really be adjusted per utterance.
The mathematical relationship IS the sound. When we made a "clean" table with quadratic closure, we literally encoded sharpness ā 2 into it. The synth doesn't care if that curve comes from: ā¢ pow(1-t, 2.0) computed live ā¢ A lookup table with those same values Same curve = same harmonics = same sound. The sample rate scaling insight: At higher sample rates, you need MORE harmonics to fill the spectrum. Lower sharpness = fewer harmonics = sounds "breathy" because there's less energy up high. That's why LF master scales sharpness UP with sample rate - to maintain consistent harmonic richness. The downside to a fixed glottal table is that you would need a different glottal table for each sample rate, if you wanted to keep the smoothness alive. By making one single Glottal table, we are effectively baking in a sharpness 2.0 into all sample rates.
I do wonder if folks liked it when the voice was smoother at 16000/22050 or 44100, or when it now has more sharpness. The sharpness is actually what Eloquence also has, when you think about it. It's voice isn't smoother. So now, with the recent SpeechPlayer, it's a bit like how the original sharpness was with it, when before we changed to the new Cosign waveform. It made 11025 sound very similar to 16000 and 22050 which is why I changed it. I didn't get any revolts, or like, "eww, where's the smoothness?" if anything one person told me that the voice sounds more "whole" at 44100. So I'm very conflicted on this, and I suspect people have their own preference. But now I am closer to understanding what gives Eloquence that classic Eloquence sharpness, they too probably also use an LF-type waveform. many blind users who grew up on DECtalk/Eloquence found the softer voices harder to understand at speed, not easier. The softer voice can make consonants stick out more and not sound like it's part of the voice's bandwidth itself.
So the sunday times hsa announced who's paying the most tax. BBC have reported on it. Concentrating on someone I'm presuming is a celebrity. A google suggests Harry Styles has a networth something like Ā£225m. A lot, but not enough to be a billionaire. In comparison there are 156 billionaires in the UK. How then, is someone with a worth of under a quarter of a bn in the top 100, when the list should basically be 100 billionaires.
I think I'm going to add a new "patrick" voice to Speech Player though, modeled after the other parameters, like F3 multipliers, extracted from the voice. It's not all lost. It's just that Glottal shape mostly impacts smoothness, hollowness, and ringyness of the voice or whether it's metallic or not. Badly tuned glottal tables get you a mess, well tuned ones should sound closer to what the synth used to sound like in the first place.
Is your Mastodon timeline too empty or too full? Are you seeing too much of one thing while missing important posts? Is your timeline overwhelmed by someone who shares too many posts?
These and lots more timeline problems are solved in the "How to customise your Mastodon timeline" guide:
So, huge thanks to @BorrisInABox for sitting down and making silly sounds for like 25 minutes for me. This is a recording of SpeechPlayer using his Glottl tables. Notice how much better it came out compared to my own, no sharpness, no boxy sound. It's a perfect glottal source. But also, this tells me that the math-derived LF glottal source with tunable sharpness could get us almost this same voice, although there's a richness in the bass harmonics I can't quite explain, so the real-source Glottal tables still do a better job at making it sound a bit more natural.
Briš„°
in reply to Mike Wassel • • •