My profession photo taken several years ago. I'm wearing a light sweater there, I'm slightly smiling.

André Polykanine

@menelion@caneandable.social · Joined ⁨Aug⁩ ⁨2026⁩

A married and happy geek from Strasbourg, France.
Backend software engineer and accessibility specialist with 20 years of experience.
I live with my wife and our two cats, and I believe cats are the cutest, yet the most independent companions.
Pagan, worshiper of Scandinavian gods.
This account is multilingual, posts may contain strong language.
No alt text, no boost.
AI is our assistant, not our replacement, not our enemy.
Pro-European, pro-American, leaning to the right.
Opinions are my own.

Spoken languages
English, Français, Українська, Русский, Deutsch

André Polykanine boosted

"fuck paypal fuck paypal" until you need someone from across the ocean to pay for a commission


can we all kind of understand that unfortunately leaving paypal behind entirely is not possible because there's no decent alternative that works around the fucking globe

the americans are still stuck in the stone age of banking and are too lazy to modernize it so it could, for example, use a fucking IBAN number to ensure *some* compatibility with other countries

all i'm saying is unless you come up with a better solution and get everyone to adopt it, businesses and individuals alike, please stop making using paypal into a moral judgement of some sort

André Polykanine boosted

A student of mine is testing audio feedback for digital whiteboards. Call for testers in this thread — boosts welcome!

For my Bachelor’s thesis, I am investigating how the use of audio feedback can improve interaction with digital whiteboards. The goal is to make digital whiteboards more intuitive and accessible and to explore which forms of audio feedback can best support users in understanding and navigating events on a whiteboard.

André Polykanine boosted

Alcohol drink recipe

I present to you a new drink recipe: Hayrish. Basically #Irish #coffee, but with #Armenian twist.

• Take 150-200 ml of home-brewed coffee
• Take 20-25 ml of Armenian cognac
• Shake these until full diffusion
• Alternatively, just pour them into one cup—but that’ll be less drinkable

The result is coffee with slight spirit-y and chocolate-y flavor. Kinda nice, compared to much more raspy Irish coffee. Or maybe that’s me liking cognac less than whisky?

André Polykanine boosted

It's honestly depressing how bad stock TTS APIs are across the board. Paperback on mobile has truly shown me this; I outline how the iOS side of this works in the quoted post, and Android is its own flavor of hacky: we synthesize ahead to temporary WAV files and chain MediaPlayer instances with setNextMediaPlayer() to kill the inter-paragraph silence put there with the stock player. On both platforms, getting a seamless reading experience means treating the system TTS purely as a text-to-speech renderer and building the player yourself.

Quin@TheQuinbox@dragonscave.space

@simon Spot on! Instead of letting iOS handle playback directly, I use AVSpeechSynthesizer purely as a speech renderer writing raw PCM audio buffers into memory, which then feed into a custom AVAudioEngine pipeline, which aggressively prefetches multiple paragraphs ahead (and behind for instant rewinds) to survive iOS throttling background synthesis when the screen is locked, resamples them to the hardware rate, and pre-arms the next buffer directly onto the audio node while the current one is still playing. That last one is also why transport controls like pausing and resuming mid-word feel instant. Say what you will about Voice Dream, but at least prior to it getting bought, have hardly used it since, it was the only app I ever saw get this right, and I wanted Paperback to as well. Android testing on your part would definitely be nice, but both of the primary Paperback developers use android as our daily devices, so that one has gotten more testing by nature of us relying on it.

André Polykanine boosted

Есть 2 интересные вещи про песню группы Девятый район "Не забывай":
1. Мне одному слышится манера исполнения Самойловых?
Когда я впервые услышал эту песню до того, как посмотреть что это Алексей Никитин и группа Девятый район я подумал, что это ну уж очень похоже на Самойловых и может это что-то сольное от одного из них.
2. Если послушать или вспомнить эту песню хотя бы раз за день, можно не переставать её петь целый день! Так и делаю периодически сегодня!
youtube.com/watch?v=CNCjocdDeuA

André Polykanine boosted

That is fascinating for me how Eloquence is so much loved and supported by community and it always will be. Interesting that it is not DecTalk but Eloquence. When I see again and again how much this TTS is loved I always think is there any russian speech synthesizer that has that much care and love of community through decades? And I think it is Newfon. Now we have Newfon, Ru-tts which is supported and maintained by Igor Poretsky together with Rulex dictionary. Newfon is also updated and maintained by different developers. We now have Newfon for iOS which I use as russian synthesizer. That's pity there was no request for Russian in Eloquence back when it was developed.

Amir@amir@dragonscave.space

Every time I think the Eloquence/IBMTTS Community Dictionary has managed to cover every possible kind of word, another batch arrives and proves me wrong.
The upcoming 2026.09 release has become even more eclectic since my last look. Alongside obscure historical and regional vocabulary such as wanhope, bealach, shaliach, and tzibur, there are names like Cécile, Daymond, Vladikavkaz, Stellenbosch, Ingushetiya, and Prakash. And then there are words that seem to have been invented specifically to make pronunciation dictionaries work overtime.
Mathematics and scientific terminology get plenty of attention this time around. Diffeomorphism, heautomorphism, heautophany, katamorphism, symmorphism, and transmorphism now have proper pronunciation support, while biology contributes an impressive collection including guanophore, katharophore, scolopophore, spermaphore, zoanthid, and zoanthus. Chemistry and physics join in with endothermicity, exothermicity, monatomicity, and a selection of wonderfully Latin looking anatomical terms such as abomasus, membranosus, nervosus, and tendinosus.
There is also a distinctly human side to the list. Fubu, twatwaffle, necromancing, necromantist, necromass, and romantist sit comfortably alongside much more respectable vocabulary, proving once again that a pronunciation dictionary has absolutely no interest in judging its subjects.
And of course, the "max" family is still growing. alcoholmax, ballmax, bossmax, confidencemax, couplemax, datemax, facemax, gymmax, pianomax, studymax, whimsymax, and many others now give Eloquence a way to deal with a whole family of rapidly evolving internet expressions, including their "maxxing" forms.
There is one more thing worth acknowledging here: the Oxford English Dictionary. Its American English pronunciations, along with its regular updates, provide an invaluable foundation for a large part of this work. Without the OED continually documenting how words are actually pronounced and how the language evolves, many of these pronunciation fixes would be considerably harder to make with confidence.
And that is perhaps the most impressive thing about the Community Dictionary. It is not merely fixing a list of mispronunciations. It is continuously extending the linguistic lifespan of Eloquence, allowing a decades old voice to keep up with words from fields and cultures that its original designers could never have anticipated.
That is a pretty remarkable second life for an old but popular TTS engine.
github.com/eigencrow/IBMTTSDic

André Polykanine boosted

This song has been haunting me for a long while now and I had no idea what it is. A Greek friend of mine helped me to identify it though but since I haven't found any decent versions yet, I decided to experiment with Suno. You can listen to the result here.
audiopub.site/listen/5992977a-

#music #suno #greek

audiopub.site/listen/5992977a-0090-4eb6-8b62-cbd627cff2a6

André Polykanine boosted

I discovered a research paper, I'll try to implement what they demonstrated, and then experiment to make it use as less resource as possible

> DuplexCascade: Full-Duplex Speech-to-Speech Dialogue with VAD-Free Cascaded ASR–LLM–TTS Pipeline and Micro-Turn Optimization
> arxiv.org/html/2603.09180

Would be fun to have a full duplex with a local LLM :D

#ai #learning #LLM #SLM #LocalLLM

André Polykanine boosted

This is an amazing program. It should appeal to anyone who produces their own radio show or podcast.

Tony Gebhard@tonygeb23@techhub.social

New and free: TG Drop Deck.
A soundboard where every sound is one keypress away. 1 to 0 fires effects, Ctrl fires your own drops, Alt+Ctrl runs looping music beds — and the beds duck themselves out of the way whenever a sound plays.
Keyboard-first and screen-reader-first, because I use it on my own show.
40 sounds and beds included, all AI generated, so they're mine to give away. Make your own decks and have fun.
tgstudios.app/drop-deck/
#Blind #Accessibility #ScreenReader #NVDA #Podcasting #A11y

tgstudios.app/drop-deck/

André Polykanine boosted

Holy crap, turns out there is actually a very big difference between Claude Opus on Pro and Claude Opus on Max. Hot damn, I am actually impressed with Opus on Max, but on Pro it just wasted time. I guess that makes sense from a business perspective, but I wrote off Opus way too quick. I doubt I'll end up keeping Claude Max, I really do not need the 5x usage limits, but figured I'd try it for a month, see how well it works and what Fable's all about etc.

André Polykanine boosted

And Paperback for iOS is now in public Test Flight! My original plan was to publish both Android and iOS at the same time, but Google is unexpectedly dragging their feet, and when I heard this email come in I literally stopped my work for a minute to come post this. It's still not perfect, there's one bug I've found so far if your book contains < or >, but that's easily fixable and now this way many more actual iOS users can throw rocks at it. testflight.apple.com/join/e562

Open source, politics

Recently I've read that the open source movement is political by definition. So to put it boldly, if you develop something open source, you are a communist, marxist, anarchist and whoever else.
Hell no.
I personally know at least one maintainer of a significant number of popular open source projects who is, to say the least, very far from those ideas politically (it's not me, I'm not so much in the open source recently, because lack of time).
So again: hell no.
#OpenSource #Politics