I could've sworn I pushed this a year ago but apparently I'd been sitting on it so...
https://pypi.org/project/rmectl
Python module to control the RME Digiface. Low effort, but it works. In case anyone finds it useful ^^
Remote
Hoshino Lina (星乃リナ) 🩵 3D Yuri Wedding prep!!!
@lina@vt.social
mastodon 4.5.18+glitch.vtsocialHacker & Mad Scientist VTuber!
💍 @cyan@vt.social 🩵 | We are @TokyoHackerGirls@vt.social ✨
ママ : https://bsky.app/profile/heripiro.bsky.social | パパ : https://twitter.com/otozuki_teru
All the links: https://lina.yt/me
16243 Followers
144 Following
50 Posts
Joined November 07, 2022
YouTube:
Twitter:
Bluesky:
GitHub:
Sponsor Me!:
Buy me a Ko-Fi!:
STREAM: Goodbye VTube Studio?!
✨ First look at OpenVT! ✨
OpenVT is a fully open source VTubing app using Ayagami by erodozer!
Let's take a look and see how it works... and take over the VTubing world!
🕐 Sep 13 13:00 JST / 04:00 UTC
▶️ https://youtube.com/live/icqDLnW5Pzk
#VTuber #ENVTuber #Live2D
Goodbye VTube Studio?! First look at OpenVT!
So let me get this straight... rsync made a *security release* fixing a bunch of CVEs, it regressed some stuff, people looked at the commit log, saw Claude sign-offs, and started a mob on the sole maintainer?
Yeah, this stuff is what gives legitimate AI criticism a bad name.
I don't like it, you may not like it either, but when people are throwing LLMs at legacy codebases and finding CVEs by the dozen, and a sole maintainer is trying to keep the house from falling apart... if you're attacking them, you're firing at the wrong person.
You know what's a bigger cancer on this world than AI? People incapable of seeing any nuance in situations. And this applies to absolutely everything. I'm absolutely exhausted of extremist takes. From every single side and point of view, in every single debate, AI related and not.
You all seriously need to touch grass, and learn to stop being outraged all the time over every single thing in this world.
Remember how I said Live2D proprietary source code is used for AI training?
Now we have proof.
Hugging Face say they exclude non-permissive licensed code from their data sets. That is a lie. They have proprietary, highly restricted code in there.
This is why Ayagami bans LLMs.
Link: https://huggingface.co/spaces/HuggingFaceCode/in-the-stack
Looks like Hugging Face scraped ~all of GitHub into an AI training dataset...
https://huggingface.co/spaces/HuggingFaceCode/in-the-stack
There is an opt-out link. You might want to use it.
✨✨ I just published the Ayagami Live2D model renderer web demo! ✨✨
https://demo.ayagami.dev/
This is a 100% LOCAL webapp, your model never leaves your computer.
Please let me know if it works for your model! (note: no physics support yet, but you can pose the physics parameters)
Source code coming soon~ 🔥
I worked at a fairly big tech co years before the AI boom. People did large scale refractoring across huge code bases back then. With refactoring tools. And properly written robots.
Applying changes to code at scale, opening PRs automatically, basic interaction with human reviewers, making sure tests pass, getting things merged when ready. All that already existed before LLMs. And it was actually reliable and not capable of hallucinating terrible things.
It's like we've forgotten how to automate things without LLMs and openclaw now...
What if Live2D rigging looked like this?
This is not a mockup. It's real. Azxiana is building a modern CMO3 editor from scratch.
Boosted by @GroupNebula563@mastodon.social
Replying to
The source code repo is up!! 🚀
https://github.com/AyagamiDev/ayagami
I've seen some people claim that AI has been used for coding for 10 years or more.
GitHub Copilot came out 5 years ago. Before that, there was nothing significant or widespread.
We can all have our opinions on LLMs, but let's not forget this is all VERY recent!
RE: @lina@vt.social
I got the Live2D physics all figured out!! As far as I can tell I have a reference Python implementation that is exact to within float precision error.
Now I need to rewrite it as a properly designed physics solver with behavior that is "close enough" without the bugs...
OMFG I just solved yesterday's Live2D physics mystery.
The trick was to stop trying to make sense of the numbers geometrically, and just try writing buggy code.
Wrote the dumbest bug I could think of, matched Live2D behavior 100% on the first try.
Spot the bug...
STARTING IN 5 MINUTES!!!!!!!! 🚀🚀🚀🚀
https://lina.yt/science
Replying to
- For your safety and privacy, there is an HTML CSP meta that blocks ALL script requests so no data can be sent anywhere at all
- This is hosted on GitHub Pages so I don't even have request log access
- I appreciate screenshots/feedback, but please be respectful of your artist/rigger! If it's broken in an ugly/scary way, a detailed description of how it's broken is enough
- DM me for private feedback!
Replying to
@iamada@tech.lgbt Ah but you see, it's not piracy if it's FOSS, because all FOSS code is effectively in the public domain.
At least that's what all the AI people think, as far as I can tell.
What is happening......???
🕗 Jul 31 @ 20:00 PDT / 23:00 EDT
🕛 Aug 1 @ 12:00 JST / 03:00 UTC
▶️ https://lina.yt/science
The Ayagami Godot extension is cooking!!! Live2D in Godot, 100% open source!
https://bsky.app/profile/erodozer.moe/post/3mqq7ckwhnk2k
Replying to
The slop machine even hard coded the exact same camera matrix from my example dumps, down to the fp64 rounding error (400 is actually a result of a calculation but the slop machine isn't smart enough to guess what it is).
(Also the math here is wrong, the slop version's entire take on resolutions and aspect ratios is nonsense)
Mystery encryption key: p3s6v9y$B&E)H@Mc
Any idea where this comes from? There are a few web hits, including this list of similarly-shaped strings that includes it: https://pastebin.com/0UTJYhBA
It feels like it's some kind of poor rand() LCG-ish thing trivially converted to ASCII, but I'm really curious if anyone has a good guess.
A certain popular VTuber software uses it as an XOR pad obfuscation for its config files... (I derived it statistically, didn't look at the code)
Replying to
BTW, that was a debug build without optimizations!! With a release build it's solid 60FPS and less than 1/3 of a core of CPU usage ^^
Replying to
@whitequark@social.treehouse.systems My nuanced position is that quitting accomplishes precisely nothing because the lack of manpower isn't going to magically resolve itself, and if you think comparing throwing Claude at a test suite to modernize it to what Bun did with the Rust rewrite is reasonable, you need a serious reality check.
If you don't like the guy resorting to LLMs to handle the deluge of LLM-generated CVEs, go help him yourself. I'll wait.
STREAM: Reverse engineering Live2D Physics! (part 2)
The Live2D physics engine is so cursed!!!! Can I get the rest of it figured out???
🕐 Jul 26 13:00 JST / 04:00 UTC
▶️ https://youtube.com/live/pbQApbjga_A
#VTuber #ENVTuber #Live2D
I was browsing @acegikmo@mastodon.social's media tab on bluesky to find something and my god why is everything she posts so PRETTY????
Every single PoC and visualization she posts could be a figure in a high quality textbook or video.
Even the random photos of her desk and cat look professional!!!
I want to be like that ;;
Replying to
OK, I have a generator. It's a float-based algorithm, so there's rounding error issues, but this script works for the above 1000-key dump, and for a contiguous chunk of another pastebin I found with longer keys.
https://gist.github.com/hoshinolina/7630ca303bcd649b8222e9abd8920caf
@dramforever@mastodon.social
RE: @lina@vt.social
Starting in 15 minutes!!! Let's get the physics wrapped up!!! ✨
Working on a little something maybe 👀
There is Discourse™ on birdsite about whether riggers should share cmo3 files, and this is my take on it.
https://xcancel.com/i/status/2060992033049976838
https://xcancel.com/i/status/2061112845870461353
TL;DR this is all Live2D Inc's fault, and no other creative ecosystem/discipline has this problem.
RE: https://vt.social/@lina/116946986988044707
Starting now!!!! Let's get physics figured out!!!
Replying to
Apparently I was so annoyed I didn't realize I typoed Swift as Switch... ^^;;
Replying to
@amarioguy@social.treehouse.systems @gfaster@tilde.zone
It's not even really "stolen", like if someone had done this a month later I couldn't care less, whatever. People will slop. It's also completely valid to take your time and do it by hand and while credit would be nice it's not required, I don't have a copyright on Apple's API.
It's this thing where AI can spit out MVPs with zero effort and time and so it allows these people to jump ahead of the person who had the idea in the first place, while investing zero time and learning nothing in the process.
Another example: https://bsky.app/profile/freya.bsky.social/post/3mhdsjgf7cc2h @acegikmo@mastodon.social
Replying to
@nina_kali_nina@tech.lgbt @moriel@chaosfem.tw @gnuplusknoppers@troet.cafe @0xabad1dea@infosec.exchange The "free software" crowd comes to mind, you know, the FSF purists buying 10 year old laptops because they're "100% free" while their home has hundreds of microcontrollers running proprietary firmware, in everything from dishwashers to their computer mouse to the EC in their supposedly 100% free laptop. They invent arbitrary rules like "ROM doesn't count" that actually go against the goals and spirit of their supposed moral values.
It's a farce and I can't take those people seriously, nor anyone who seriously believes "zero AI involved in my software" is a viable thing going forward.
Go support projects that don't embrace AI and stop worrying about the supply chain, because that's a lost cause and acting like it isn't is a losing battle that leads to madness.
Replying to
@nina_kali_nina@tech.lgbt @moriel@chaosfem.tw @gnuplusknoppers@troet.cafe @0xabad1dea@infosec.exchange It is a choice if you are willing to make extreme trade offs, that no existing established project the size of Rust can make and survive.
You as an individual are free to make whatever choices you choose, but I do hope those choices are at least self consistent, because 90+% of the time the people complaining about this stuff and saying they're going to do something about it are just virtue signaling and then not actually willing to make all the compromises required to maintain an internally consistent position.
To be clear, I don't like any of this AI slop stuff either, but I'm not going to switch to NetBSD because of it, and the tiny minority of people who are aren't going to make an impact on the field.
Software is not like food, it's not a consumable that is continuously created in a cycle. Refusing to use code touched by AI is like a vegan refusing to eat food that was, at any point in its life cycle, produced or touched by a non vegan or had any item, device, or tool designed or manufactured by a non vegan used. Is it possible? Yeah, if you go live off grid entirely off of natural resources and make your own tools to harvest food. But I highly doubt most vegans would be willing to do that.
Replying to
@linuxandyarn@hachyderm.io @unlambda@hachyderm.io @dramforever@mastodon.social More importantly, there are a zillion password generator websites that generate the password *locally* in JS, which makes them infinitely more trustable than what DDG is doing.
I would go as far as saying that DDG feature is appalling, and should not exist as implemented. It makes me want to stop using it as my default search engine. No responsible website should ever even attempt to implement a password generator like that. That's exactly what brought us to the mess I mentioned in OP.
Replying to
@linuxandyarn@hachyderm.io @dramforever@mastodon.social
...what. Why would you trust duckduckgo with server-side password generation?!
If you're a VTuber or rigger and you have "interesting" experiences with Live2D physics I'd love to hear from you!
Are you bothered by the quirks? Do you "abuse" them for a model? How do you configure physics FPS in VTS?
I want to make sure my version does what users want!
Replying to
@sabik@rants.au @whitequark@social.treehouse.systems The vast, vast, vast majority of the recent changelog co-authored-by Claude is test suite work.
But people see a page of commits with "claude" in the author tagline and freak out, completely ignoring that almost everything is "testsuite:" or "ci:".
Replying to
@moriel@chaosfem.tw @nina_kali_nina@tech.lgbt @gnuplusknoppers@troet.cafe @0xabad1dea@infosec.exchange If the goal is "no AI generated code in my dependency tree", then that was a battle lost before it even began. Multiple core dependency projects are allowing AI.
Given the state of the world re AI usage, there was no way that battle was going to be won. You'd need something like "AI is ruled illegal/copyright infringing across the board in some important jurisdiction" for people to consider it toxic enough to not touch and fork each and every project whose maintenance team disagrees.
The way FOSS is interconnected, it is simply not possible to "hard partition" the dependency tree on moral concerns at scale. Pretty much the only thing that can do that is a legal code red, and it seems the lawyers have decided AI isn't one (even though many disagree).
Put another way: The only entities with the power to credibly fork projects on a whim to uphold requirements of this class are big corporations (see: OpenSearch), and big corporations are overwhelmingly pro AI.
There's some talk of Linux distros disallowing AI in packages, but even if something like Alpine were to pull that trigger, I highly suspect one or more FAANGs would just fork Alpine in response...
Replying to
@ahltorp@mastodon.nu And compilers also had AST parsers that could be used as part of refactoring pipelines!
Replying to
@Profpatsch@mastodon.xyz @whitequark@social.treehouse.systems Because leftist infighting promoted by people who must ensure every person around them shares their precise views or else they are Bad is all the rage these days. Haven't you heard? The AI people (and MAGAs) love it, it means their enemies self implode while the peolple causing it think they're righteous and innocent and only doing good for the world.
Replying to
@linuxandyarn@hachyderm.io @unlambda@hachyderm.io @dramforever@mastodon.social Without doing a full audit, the top 5 results on DDG for "password generator" look local, judging by latency (checking on phone, too lazy to pull up DevTools on a PC and check properly).
It really is irresponsible to implement a password generator site in any other way.
Replying to
@ChuckMcManis@chaos.social @sabik@rants.au I saw the posts. It was literally "I saw Claude tagged commits, shock and horror".
Replying to
@moriel@chaosfem.tw @nina_kali_nina@tech.lgbt @gnuplusknoppers@troet.cafe @0xabad1dea@infosec.exchange So what you're saying is "it isn't possible to be 100% pure but I'm going to choose what programming language to learn based on some rando on social media pointing out a dependency that allows AI".
Good luck with your ethics if that's what you base them on... I'd rather make ethical choices based on realistic, informed positions I can take that are meaningful, rather than whatever is cool to virtue signal this week.
Replying to
@emc2@indieweb.social Small local LLMs can do that too. I don't know why nobody is working on those more...
Replying to
@kfogel@kfogel.org My argument does not require any system of copyright beyond basic author's moral rights. Since LLMs can reproduce copyrighted works, it follows that the copyrighted work is embedded in the LLM. Any interpretation that dismisses this to the extent the original author would have no interest nor control in their work when it is in the form of LLM/AI output (or any intermediate form capable of producing such output, or any output that is derivative, which LLMs can also do) would require essentially complete abolition of copyright as we know it, or otherwise would become a backdoor capable of negating the effect of any system of copyright.
While endless discussions could be had about the flaws of copyright legislation as it exists today, I don't think you'll find that "copyright shouldn't exist at all" is a very popular take, especially among creative fields.
Should you wish to make an argument against copyright-based opposition to GenAI, I think you'll have to present a workable and popularizable alternate system of copyright first to have a credible argument.
Either that or get UBI rolled out first, but good luck with that when the billionaire class is busy taking the world's money to deploy and run AI. And those billionaires are, morally, the furthest thing from jazz musicians.
Cyan: *Holding a book* what do you call the thing that marks a page, like a bookmark?
Me: ... a bookmark
Cyan: No like a physical one
Me: Where do you think the word comes from?!?!?!?!?!?!
(This seriously just happened)
Replying to
@kfogel@kfogel.org TL;DR if you want to argue copyright shouldn't be the foundation of an ethics argument, you're going to have to first build up the ethics of copying creative works from first principles, because "it's always ethical to copy" isn't going to get a lot of people on your side.