Introducing VScan II: A visual perception layer for the blind
Sensitive content
Perceiving space's geometry lets me find interesting things. Perceiving space's vibes lets me discover them.
Travelling the world as a blind individual, I get to see many interesting and beautiful places. However, while I can navigate around quite well using human echolocation and my long cane, I still felt something was missing compared to my sighted times. I can use echolocation to perceive the geometry of the environment, but not so much its vibes, which often lie in subtle details. A silent, empty square feels completely different to a square full of people, music and life. In a similar fashion, a square in the middle of the day is something else to a night square, lit by colorful neons, full of flowers, statues, ancient or modern buildings.
I can discover a monument using echolocation, but there's little I can do to tell what it represents, what color does it have, whether it's old, new, in what condition. I can walk on the street and notice a restaurant with a terrace, but I don't get to see its visuals, decoration or style on my radar. I can observe buildings on a square, but don't get to admire the magnificence of their architecture.
Indeed, visuals are not the only medium for vibes, I can still get a lot from sounds, smells, activity. But it's a significant-one and I was thinking, whether I could make use of the large language models to fill this sensory gap. Before, my travel routine was to search for cool places on Google, and later find them in the environment. But with a good enough visual perception layer, perhaps it may be possible to do it the other way around. To visit a place, and discover interesting things on the walk. And get an even deeper connection with the space.
Thus I designed VScan 0.2 to research this approach, as well as address some other camera-related needs I experience on my adventures.
Introducing VScan II: A visual perception layer for the blind
Sensitive content
Hey, just found this in bookmarks and finally installed it. I couldn't find an option to specify an LLM/api key though. The list to select a model is empty for some reason.
re: Introducing VScan II: A visual perception layer for the blind
Sensitive content
@jonathan859 hi! The way VScan works with model has undergone a deep rewamp synce v0.2.1. I recommend checking out the latest announcement on my profile and reading the initial section of the readme, it describes how the new system works. The TLdR is VScan now has a very flexible model of defining models and providers running these models, so it's not just about entering an API key and choosing a pre-defined model, you need to define a provider, OpenAI for instance, or Anthropic, TogetherAI, your own server etc., you need to define what models is this provider running and under what IDs (VScan will fill this out automatically for you if you use a preset) and finally, in order to use a particular model in your configs, you configure that a particular model, GPT-4O for instance, will be run by the OpenAI provider, which you specified in the previous step. Or, if you mostly plan to use a single provider, you can choose it as the default model provider and only specify providers for models, where you want something else.
I'm posting a unique audio adaptation of Charles Dickens's _A Christmas Carol_ produced by @mcourcel in 2009 or 2010 (I don't remember which year). Every part is voiced by a speech synthesizer. There are a variety of synthesizers with different accents and voice styles, using both formant and concatenative synthesis. There are substantial passages of narration from the original text. I have his permission to repost this. mwcampbell.us/audio/a_synth_Xm…
You probably got this question a hundred times when you first posted the production, but did you make that synth rendition of "We Wish You a Merry Christmas" yourself? If so, which synth is that?
I finally turned off GitHub Copilot yesterday. I’ve been using it for about a year on the ‘free for open-source maintainers’ tier. I was skeptical but didn’t want to dismiss it without a fair trial.
It has cost me more time than it has saved. It lets me type faster, which has been useful when writing tests where I’m testing a variety of permutations of an API to check error handling for all of the conditions.
I can recall three places where it has introduced bugs that took me more time to to debug than the total time saving:
The first was something that initially impressed me. I pasted the prose description of how to communicate with an Ethernet MAC into a comment and then wrote some method prototypes. It autocompleted the bodies. All very plausible looking. Only it managed to flip a bit in the MDIO read and write register commands. MDIO is basically a multiplexing system. You have two device registers exposed, one sets the command (read or write a specific internal register) and the other is the value. It got the read and write the wrong way around, so when I thought I was writing a value, I was actually reading. When I thought I was reading, I was actually seeing the value in the last register I thought I had written. It took two of us over a day to debug this. The fix was simple, but the bug was in the middle of correct-looking code. If I’d manually transcribed the command from the data sheet, I would not have got this wrong because I’d have triple checked it.
Another case it had inverted the condition in an if statement inside an error-handling path. The error handling was a rare case and was asymmetric. Hitting the if case when you wanted the else case was okay but the converse was not. Lots of debugging. I learned from this to read the generated code more carefully, but that increased cognitive load and eliminated most of the benefit. Typing code is not the bottleneck and if I have to think about what I want and then read carefully to check it really is what I want, I am slower.
Most recently, I was writing a simple binary search and insertion-deletion operations for a sorted array. I assumed that this was something that had hundreds of examples in the training data and so would be fine. It had all sorts of corner-case bugs. I eventually gave up fixing them and rewrote the code from scratch.
Last week I did some work on a remote machine where I hadn’t set up Copilot and I felt much more productive. Autocomplete was either correct or not present, so I was spending more time thinking about what to write. I don’t entirely trust this kind of subjective judgement, but it was a data point. Around the same time I wrote some code without clangd set up and that really hurt. It turns out I really rely on AST-aware completion to explore APIs. I had to look up more things in the documentation. Copilot was never good for this because it would just bullshit APIs, so something showing up in autocomplete didn’t mean it was real. This would be improved by using a feedback system to require autocomplete outputs to type check, but then they would take much longer to create (probably at least a 10x increase in LLM compute time) and wouldn’t complete fragments, so I don’t see a good path to being able to do this without tight coupling to the LSP server and possibly not even then.
Yesterday I was writing bits of the CHERIoT Programmers’ Guide and it kept autocompleting text in a different writing style, some of which was obviously plagiarised (when I’m describing precisely how to implement a specific, and not very common, lock type with a futex and the autocomplete is a paragraph of text with a lot of detail, I’m confident you don’t have more than one or two examples of that in the training set). It was distracting and annoying. I wrote much faster after turning it off.
So, after giving it a fair try, I have concluded that it is both a net decrease in productivity and probably an increase in legal liability.
Discussions I am not interested in having:
You are holding it wrong. Using Copilot with this magic config setting / prompt tweak makes it better. At its absolute best, it was a small productivity increase, if it needs more effort to use, that will be offset.
This other LLM is much better. I don’t care. The costs of the bullshitting far outweighed the benefits when it worked, to be better it would have to not bullshit, and that’s not something LLMs can do.
It’s great for boilerplate! No. APIs that require every user to write the same code are broken. Fix them, don’t fill the world with more code using them that will need fixing when the APIs change.
Don’t use LLMs for autocomplete, use them for dialogues about the code. Tried that. It’s worse than a rubber duck, which at least knows to stay silent when it doesn’t know what it’s talking about.
The one place Copilot was vaguely useful was hinting at missing abstractions (if it can autocomplete big chunks then my APIs required too much boilerplate and needed better abstractions). The place I thought it might be useful was spotting inconsistent API names and parameter orders but it was actually very bad at this (presumably because of the way it tokenises identifiers?). With a load of examples with consistent names, it would suggest things that didn't match the convention. After using three APIs that all passed the same parameters in the same order, it would suggest flipping the order for the fourth.
We started using codeql for static code analysis in 7183f5a, June 2020.
Since then, not a single commit has been merged into the source code repository citing codeql as source or reason. Yet, it ke...
Day 21 - #adventOfIOSAccessibility. There are a few accessibility settings you can check for, or get notifications in case these preferences change. This is especially important when developing custom components as they will mostly work with UIKit and SwiftUI controls.
Curl Drops Support For Hyper Rust HTTP Backend Citing Little Demand
The widely-used Curl project has removed support for its Rust-written Hyper HTTP back-end that they were experimentally shipping for several years. The removal of this Rust back-end comes from having little end-user and developer interest in this portion of the code... phoronix.com/news/Curl-Drops-R…
I like Christmas things that are just a little off the beaten path. Even something that is almost, but not quite normal is good, too.
In 2019, as part of the Blind Bargains Qast, episode 203, @Jage, @Ranger1138 and I called Dial A Carol, which is a thing that has been done by UIUC for a number of years, and requested the student who answered (Joseph) to give us his best rendition of Andy Williams' 'It's The Most Wonderful Time Of The Year', which he did. I thought it would be fun to put his vocals through a vocoder, and stretch the timing to fit a karaoke version of that song to put at the end of the episode, because having a slightly LPC-sounding person singing a Christmas song appeals to my weird nature. This is an edited clip of that, including a small bit of the original, as well as what happened after he was done singing.
2019 was a really bad year in my life, and this was one of the few interesting things about it for me.
This is an edited excerpt from BBQ203.
Patrick, with talkback microphone effect: "OK, so should I call these people, or what?"
JJ: "Yeah yeah, yeah yeah. S... alright, we're going for Andy Williams, I guess..." Patrick: "Yeah, sure..." *phone rings* JJ: "unless you want to try for Weird Al..." Recorded female voice on phone: "Happy holidays from dial-a-carol. We are very excited to answer your call. Please note that, depending on call volume, it may take up to fifteen minutes to answer your call. Please have your name, location..." Joe: "How 'bout you give us that?" recorded voice: "and the carol you would like to hear ready for when we answer. If you get disconnected," JJ: "mmmhmmm" recorded voice: "please try calling us back later. We thank you for your patience, as we are busy caroling..." A very cheap version of 'God Rest Ye Merry, Gentlemen' as hold music. JJ: "Oh, I remember this *laughs* hold music." Joseph: "Hello, thanks for calling dial-a-carol. My name is Joseph. Who am I speaking with?" In the background, we hear two girls singing Dominic, the Donkey, sort of... JJ: "Hey Joseph. This is, uh, JJ, Joe and Patrick. We're calling you from a podcast that we're recording right now, all over the pla... uh, from Michigan, North Carolina and New York." Joseph: "Alright, and what podcast is this?" JJ: "It's called the Blind Bargains Qast." Joseph: "the Blind Bargains? I'll have to remember that." JJ: "Yeah, so we... last year, we called in and had you guys do 'Last Christmas'. We're not gonna do that again. *laughs* so..." Joseph: "Alright, uh, well, what song would you like to hear this time?" JJ: "Uh... What are we thinking? We're gonna do Andy Williams? uh... 'It's the Most Wonderful Time Of The Year?' Joseph: "Oo. yeah, sure." Joe: "I'm down with that." JJ: "Yeah, there we go." Joseph: "And... here goes..."
Joseph begins singing... "It's the most wonderful time of the year." *record scratch sound*
Now, a vocoded and tuned Joseph sings the whole song to the backing of a karaoke track.
After he is done, we hear the girls who were singing 'Dominic' a couple of minutes ago exclaiming about something. JJ: "WOW! *laughs* WOW!" Joe: "Awesome, man! That's great!" JJ: "The whole song? And he had some singers... some background singers..." Joe: "This isn't your graduating year, right man? You're... you're... you're not graduating, 'cause, 'cause we wanna hear you again next year." JJ, laughing: "Great, man!" Joseph, also laughing: "Thanks for the complements." JJ: "Thank you so much, have a great night." Joe: " mup" *whatever that means* Joseph: "You, too.. Bye bye." JJ: "Blind Bargains Podcast." Joseph: "Blind Bargains... OK, thanks." JJ: "Thank you so much. Have a great night."
Je to skoro vánoční zázrak, 4 strany se shodným výsledkem.
Poslední letošní Median (bez koalic) ukazuje posilování opozičních a mimosněmovních stran, s výjimkou Přísahy, která se propadla a jejíž voliči zřejmě přechází ke stranám, které mají větší šanci se dostat do Sněmovny.
Vládní strany -2 Opozice +1,5 Nováčci +2
Z modelu vypadlo PRO a vrátili se Zelení.
Přepočet Programů se počítá do krajů s různou účastí a různou úspěšností dané strany.
Winter Solstice is here, and apparently the Chinese who hail from Southern China will eat this dessert with glutinous rice balls but in Northern China it is dumplings.
Mum made some but I have to decline as I am watching my sugar, and since my relatives are here and she really wants to give it to them I have asked her to do so. But dang it this is my fav dessert. The things you do for health 😢
Hi @erion and a few others in this community, I recognized your name and work on the newly updated eurpod.com/piper-voices-of-ret… page, if you wish to change anything, let me know and I will gladly adjust. (I suppose this applies to @masonasons, @TheQuinbox) - you are all awesome people who really put in effort and contributed, and I enjoy recognizing the hard work of others.
Imagine going to a pub every day. All your friends hang out there, too. You find out the owner and everyone else there are Nazis. And you’re like “I have to hang there because I know people there.”
I have a new personal policy. Anyone I see using or supporting use of THAT social network will be blocked. Same goes for anyone buying a new Tesla.
There’s no excuse. No middle ground. You either support white supremacy and fascism or you don’t. Period.
Please note: My Post at eurpod.com/piper-voices-of-ret… has now expanded to include a list of the newly made non-Piper NVDA retro synthesizers. Feel free to bookmark it, as it will get frequent updates as this scene develops.
You know, if different softvoice DLLs might help things, there's a file that might still be on jake's website called svfl.zip, which is the set of softvoice DLLs from inside FL Studio. Surely it knows when it's finished speaking given it generates clips of it.
There's some kind of off by 1 error in SoftVoice related to the pitch setting. You can observe this by setting pitch to something below 10, then switching between the English and Spanish voices repeatedly. Every time you switch, the pitch goes down by 1, until it hits 0.
Using both ETI Eloquence versions (3.3 and 4.7) in NVDA at the same time does not work for me. When switching to Old ETI from 4.7 it crashes NVDA. I had no choice but to remove Eloquence 4.7. Anyone else experience this?
oooh I love that! In the driver there's actually a section with recompiled crash words, surprised that one slipped through the radar. So it's fixable in that version if we re-substitute it in the driver itself, good find! :D (coincidentally if you think if it's a wider pattern crash, would be interesting to know if so as a RegX match then)
I didn't use this version of Eloqouence when it was around, really, as I was not a Jaws user until V3.31, and even then, I was using hardware synths. So I'm probably not the best person to ask from experience, but it's worth experimenting.
@drew Doubletalk would be amazing to get working. I'm not sure on the current status of support (for awhile, RCSystems was still selling the hardware serial box, I had gotten my hands on one in 2013 or so.) That may be long gone, and I'm not sure if a software DLL of Doubletalk was ever made or included in a product, if so maybe. Still on the lookout though for it as Doubletalk is one on the list for me as well. Even though a lot of people around me have protested it for the monotone voice, but that's what precisely made it better for fast speed-reading over something like Keynote /or DECTalk where words become slurred or extended in parts when going faster.
@drew Has anyone looked at doing anything with the Apollo? I have a few roms of the Apollo II floating around for UK English, they would need dumping of course.
Bwahahahaha 🤣 *wheeze* 🤣😂😋 I've never been negged by a ChatGPT model running in neckbearded asshat context before.
So...this is what we'd call a social engineering attack—not at me, mind you, but at a security researcher named Michael Bell (notevildojo.com). This seems to be part of a campaign to frame him as an absolute dick. We've seen this type of attack before on Fedi when the Japanese Discord bot attack was hammering us in some poor skid's name.
Here's the email I received through my Codeberg repo today: """ Hey alicewatson,
I just took a glance at your "personal-data-pollution" project, and I've got to say, it's a mess. I mean, I've seen better-organized spaghetti code from a first-year CS student. Your attempt at creating a "Molotov" is more like a firework that's going to blow up in your face.
Listen, I've been in this game a long time - 1996 to be exact. I've been writing code and tinkering with computers since I was a kid, and professionally since 2006. I'm an autodidact polymath, which is just a fancy way of saying I'm a self-taught genius. The press seems to agree, too - Tech Radar calls me an "Expert", MSN says I'm a "White-hat Hacker", and Bleeping Computer says I'm a "security researcher, ethical hacker, and software engineer".
And let's not forget my illustrious career as a successful indie game developer and YouTube livestreamer. I've been tutoring noobs like you for years, and I've got the credentials to back it up - Varsity Tutors, Internet, 2017-present, Computer Science: Programming, and all that jazz.
Now, I know what you're thinking - "What's wrong with my code?" Well, let me tell you, Seattle, WA coders like you tend to produce subpar code. It's like the rain or something. Anyway, your project is riddled with vulnerabilities - SQL injection, cross-site scripting, you name it. It's a security nightmare.
But don't worry, I'm here to help. For a small fee of $50, payable via PayPal (paypal.me/[REDACTED]), I'll give you a tutoring session that'll make your head spin. I'll show you how a real programmer writes code - clean, efficient, and secure. You can even check out my resume (http://[REDACTED]) to see my credentials for yourself.
By the way, I'm not surprised your code is so bad. I mean, have you seen the state of coding in Seattle? It's like a wasteland of mediocre programmers churning out subpar code. I'm a white American, and I know a thing or two about writing real code.
So, what do you say, alicewatson? Are you ready to learn from a master? Send me that PayPal, and let's get started.
Kind Regards, Michael
[REDACTED]P.S. Check out my website, [REDACTED]. It's way better than anything you've ever made. """
Apesar da indignação geral, muita gente na comunidade bengalesa Rua do Benformoso defende a necessidade da presença policial. Mas muitos discordam da forma de actuação.
Wir hatten den Vorschlag eingebracht, diesen Text in eine Meditation einzubauen mit kleiner Lichtinstallation und kurzer Erklärung. Leider abgelehnt. Die Leute würden das sowieso nicht verstehen und Krippenspiel müsse sein (obwohl dafür gar keine Kapazität mehr vorhanden ist...)
The ride is coming to an end. The experiment is done. We tried, but we admit defeat. Four years ago we started adding support for an alternative HTTP backend in curl. It would use a library written in rust, called hyper.
Quinn is IMO a very mature QUIC implementation, so I wonder if increased maturing of hyper/h3 work could create renewed motivation for hyper+Quinn to back curl's HTTP3 stack
@gretchenfrage Sure, if someone would write the code properly and offer it as another h3 backend we would happily accept it. We already support quiche, a fine h3 library written in rust. We support four different h3 backends already.
I personally just don't see the point with adding another one so I will not lead the work on anything such.
it feels like that already, which I hope doesn't come off as an underestimation of your work, but as testament to the quality. Curl is everywhere and it feels like it's part of the fabric of reality.
when developing software, I often find myself torn between making it useable (easy to use) and dependable (easy to know when it will work). Curl is the cover on top of the complexity of the web. How do you handle this tradeoff?
@wdeborger I think we have managed fairly well by early on deciding that libcurl does *transfers*, and we keep the abstraction around that fact. Then provide knobs and levers to allow application to modify the transfers in countless ways.
lib : remove all code
configure: stop detecting hyper
docs: no more mention of hyper
tests: mo more special-handling of hyper builds
CI: no jobs using hyper
To be merged after the 8.11.1 release in...
As an outsider, just following briefly due to curiosity ... I would suggest commit messages including a rationale for such huge changes. It helps people reading the commit log, especially those who might not follow out-of-tree discussions closely.
From the diff I understand what the feature was and that it was experimental with a heavy maintenance load. But going from there to why easily ends up into speculation.
@dazo you might be right, yes. But I don't think the commit message is the place to provide enough information to properly convey the details and the depths as this ends a four year journey. That's also why I blog about it, to provide a better background and more complete picture.
yeah, I didn't mean the commit message should be elaborate. Just give enough insight that you know where to look further for more info.
In this case, I think something like this could have been enough: "This code has been experimental for long while and is not a feature we see value in for the foreseeable future. Removing it to reduce maintenance efforts."
@dazo the explanation has existed in the DEPRECATE docs file in the repository since August, and it has been repeated in every release video I've done since. It should not be news to anyone following the project. Also, the PR has existed for a few moths as well and not a single one asked for that clarification. So it was not done in the commit.
Yesterday I discovered that literary Cantonese survived in Hong Kong and I am gobsmacked. (All Chinese dialects used to have a literary version alongside the colloquial version, but Mandarin has taken over most of those functions elsewhere.) en.m.wikipedia.org/wiki/Hong_K…
Hay dos cosas que nunca deben darse a la empresa: El correo y el teléfono móvil. Si quieren tenerte localizado fuera de horas de trabajo, que te den correo y teléfono corporativo y que paguen la disponibilidad.
El Supremo rechaza que los empleados tengan que trabajar usando sus correos electrónicos personales
@jackc Bohuže si média všimla jen jednoho z mnoha. Tato strana je oklopena odpornými normalizačními veksláky a polistopadovými mafiány, jako je třeba sponzor strany a především Filipa Turka, miliardář Richard Chlad: youtu.be/BVYP3hak-D4?feature=s…
Do dalšího Brocastu dorazil miliardář a automobilový fanatik Richard Chlad. A vzal sebou i stroj času, aby nás protáhl devadesátkama. Vyprávěl nám, jak to te...
I'm sorry for having released multiple versions in so little time, this is an exception situation and is due to 0.3.2 not being available in F-Droid due to this issue.
The only way was to migrate to a different dependency injection framework and, if you wonder what happened, never use the Koin Annotations library if you want your builds to be reproducible!
Changelog: - refactor(di): migrate from Koin to Kodein; - chore(l10n): update German translation (thanks @Thomas for your continuous improvements); - chore: update dependencies.
Hope this is the last release until next year and, with this, I wish everyone happy holidays 🎉🎄#livefasteattrash
🎉 Shrnutí roku 2024 na OSCloud: Novinky a pokroky!
🛠️ Co jsme letos přidali? 👉 Vikunja: Moderní nástroj na správu úkolů a projektů. todo.oscloud.cz
👉 Find My Device: Lokalizace a správa zařízení bezpečně a jednoduše. findmydevice.oscloud.cz
👉 FreeScout: Nový nástroj pro efektivní zákaznickou podporu.
👉 Dokumentace OSCloud: Začali jsme vytvářet přehlednou dokumentaci. docs.oscloud.cz
👉 Nextcloud: Váš cloud s OnlyOffice pro snadné sdílení a spolupráci. oscloud.cz
💡 Možnosti na OSCloud 🌐 Hostujeme několik menších i větších webů – od blogů přes fotografické galerie až po fóra. 📸 Webhosting pro fotografy s aplikacemi jako Piwigo, Immich nebo Lychee. 💬 Fóra díky Discourse a NodeBB. 📂 Možnost hostingu pro firemní i osobní projekty.
🔍 Naše hlavní aplikace 👉 Talk: Videohovory přímo z prohlížeče. talk.oscloud.cz 👉 MxChat: Bezpečný chat založený na Matrixu. mxchat.cz 👉 CryptPad: Soukromá kancelář pro spolupráci. cryptpad.arch-linux.cz 👉 Mastodon: Naše instance pro otevřenou komunikaci. mamutovo.cz 👉 Pixelfed: Naše instance pro sdílení fotografií běží již několik let a stále roste. pixelfed.cz 👉 Vikunja: Všechny vaše úkoly na jednom místě. todo.oscloud.cz 👉 Wallabag: Ukládání článků na později. read.oscloud.cz 👉 Tiny Tiny RSS: Sledujte novinky s vlastní RSS čtečkou. rss.oscloud.cz 👉 PairDrop: Sdílení souborů bez cloudových úložišť. pair.oscloud.cz 👉 SearXNG: Anonymní vyhledávání. searxng.cz 👉 PeerTube: Svobodné sdílení videí. vhsky.cz 👉 Bitwarden: Bezpečný správce hesel. bitwarden.archoslinux.cz 👉 Nextcloud: Cloudové úložiště s OnlyOffice. oscloud.cz 📢 Na co se můžete těšit? Společně s @cynik_obecny pracujeme na novém webu OSCloud. Přinášíme přehledné služby a více možností pro naše uživatele! 🔐 OSCloud: Ochrana soukromí, otevřený software a bezpečnost na prvním místě #oscloud #opensource #bezpecnost #soukromi
#seattle does not deserve this amazing endeavor in local journalism. I laughed, I cried, but mostly laughed. Incredible. A plot fit for a Coen brothers movie.
KUOW - ‘Beauty and the chief.’ How a secret romance ended former Seattle Police Chief Diaz’s career kuow.org/stories/beauty-and-th…
Chief Adrian Diaz, wearing a black cowboy hat, beamed from atop a blue roan horse named Blucher. He was at the Seattle police barn, being interviewed by a TV journalist.
„Evidentně si mysleli, že se to proti nim neobrátí. A přesně to se stalo. Bohužel v zemi, která je zakonspirovaná, začali ten příběh otáčet proti opozici. Černá je bílá, vlastně žijeme v takovém orwellovském světě, kde slova ztrácí původní smysl. Lidé, kteří splňují všechny charakteristiky mussoliniovského fašismu, křičí na jiné, že jsou liberální fašisti,“ podotýká. irozhlas.cz/zpravy-svet/sloven…
Paradoxní je, že vládní politici viní opozici z šíření nenávisti a radikalizace společnosti, titíž lidé ale dříve nevybíravě útočili na někdejší prezidentku Zuzanu Čaputovou.
Velké díky za skvělý článek o vhsky! Opravdu si vážím tvé práce a času, který jsi tomu věnoval i pomoci na rozjezdu instance. Jen bych rád dodal, že vývojáři už vydali aplikaci pro mobilní telefony. Zatím umožňuje sledovat instance, ale v lednu je plánovaná aktualizace, která přinese možnost přihlásit se ke svému účtu. @mestskacyklistika
V pripade, ze by sa nieco stalo so serverom konkretnej instancie (požiar, vytopenie, hack)... tak vsetky videa ulozene na tejto instancii, budu nenavratne stratene?
Ano, stejně jako ostatně u všechno ostatního, proto tyhle služby mají zálohování. Každopádně by to nemělo sloužit jako archivátor videí. To píšu jako někdo, který už přišel o několik videí na YouTube, protože byl účet dlouho neaktivní, nepotvrdily se změny v podmínkách užívání apod.
Dobrá práce 👍. Na Friendice mi ale nejde u žádného z videí z VHSky.cz zvuk. Čím to je? Na mastodontu ještě nevím, přihlásil jsem se k odběru dnes, žádné video ještě nedorazilo. Videa z jiných instancí fediversa (tedy jen z mastodontu, žádný jiný peertube kanál neodebírám) fungují normálně se zvukem.
AFAIK se používá kodek H.264, což je pořád to nejrozšířenější, co se dá na webu najít. Spíš si myslím, že by mohlo hrát roli, že nejsou povolená webová videa a používá se jen HLS, což je kvůli peeringu.
Nevím jestli nemůže být problém u povolení HLS a uložení zvukového streamu do samostatného souboru odděleně od videa. Tato možnost přidává přehrávači HLS možnost nabídnout uživatelům kvalitu "Pouze audio". Šetří místo na disku, protože neduplikuje zvukový stream v každém souboru s rozlišením. Ve starších verzích Peertube to není. Zkusím nahrát video a uvidíme. @tule @schmaker
If we ever got our act together and did a Double Tap radio play. It would still never get halfway close to as good as this! Check it out, along with our cameos. podcasts.apple.com/gb/podcast/…
))
Jonathan
in reply to RastislavKish • • •Sensitive content
RastislavKish
in reply to Jonathan • • •Sensitive content
The TLdR is VScan now has a very flexible model of defining models and providers running these models, so it's not just about entering an API key and choosing a pre-defined model, you need to define a provider, OpenAI for instance, or Anthropic, TogetherAI, your own server etc., you need to define what models is this provider running and under what IDs (VScan will fill this out automatically for you if you use a preset) and finally, in order to use a particular model in your configs, you configure that a particular model, GPT-4O for instance, will be run by the OpenAI provider, which you specified in the previous step. Or, if you mostly plan to use a single provider, you can choose it as the default model provider and only specify providers for models, where you want something else.