Peter Vágner reshared this.

RScan now on Google play! Identify products, swiftly

RScan lets you scan a barcode of any product, and automatically looks it up on DuckDuckGo and tries to select the most fitting and useful description. The scanning is really fast and convenient, meaning it's easy to work with even if you're blind and don't know the location of the barcode, but it's also great for sorting through large number of items, if you need to say search for a particular chocolate in a pile, or you order a large number of cans that you need to tell apart in order to sort them to groups. RScan can deal even with scanning multiple items of the same barcode, making this process easy and efficient.

I've been gradually developing this app over the last years, always reflecting my actual needs and actively using it in my home. I've been sharing it with my friends from Czechoslovakia, receiving great feedback. In our region, the automatic product identification works really well and RScan can truly identify 90% of items just by seeing their barcode.

Few days ago, I finally got to release it in google play, and given this opportunity, would like to expand it to more regions. My theory is that my algorithm should work comparably well in regions where products use metric system for describing themselves (300 g chocolate, 400 ml can etc.).
It's not yet prepared for UK, where the commonly used units are different, and completely unknown for me is USA, where a similar but different barcode standard is used than in Europe, plus they have different units too.

I would love to support as many countries and regions as I can. If you would like to help me with this, sending me barcode numbers of things you commonly use in your region (cans, chocolates, common drugs, drinks), along with the name of your country would be very useful, so I could check out how does your Internet describe things and make RScan adapt for it.
If you're technically skilled, you can read the project's readme:
github.com/RastislavKish/RScan
and directly look for the unit information required by RScan, this would save me the struggle with localization and location simulation.

Also, if RScan works well in your country, I would love to hear that too! My theory of compatible countries is still just a theory, it needs to get verified.

You can find RScan on Google play:
play.google.com/store/apps/det…

And, along with all the code and scanning tips in the documentation, on my GitHub:
github.com/RastislavKish/RScan

Happy scanning!

Peter Vágner reshared this.

Peter Vágner reshared this.

I made web component out of a common pattern I use in UI, combining a range slider with a number input- potch.me/2023/range-num-web-co…

reshared this

Peter Vágner reshared this.

Are you interested in making your notifications private?
We have a great guide in our website describing how to set up your android xmpp application to deliver notifications for all your apps on your device. Without using google.
joinjabber.org/tutorials/servi…
and for people who self host you can use your own server to deliver notifications privately to your devices and the devices of the people that use your server :)
joinjabber.org/tutorials/servi…

#privacy #xmpp #dataprotection

reshared this

Peter Vágner reshared this.

This festive season give the gift of accessibility with an NV Access donation. Your support ensures we continue delivering life-changing technology to blind and visually impaired people globally.
Donations can be made here, nvaccess.org/support-us/#donat…

#Donate #Donation #NVDA #ScreenReader #Accessibility #Christmas

Peter Vágner reshared this.

Peter Vágner reshared this.

TalkBack 14.1 comes with image descriptions (which are actually surprisingly accurate from my limited testing), spell check while using the Braille keyboard, automatic scrolling for Braille displays (with a customizable speed), and (most surprisingly to me), new haptics! I'm not convinced I like them yet, just because text elements don't appear to have a vibration, but it actually feels like Voice Assistant or VoiceOver now!

reshared this

Peter Vágner reshared this.

In a matter of a couple of weeks, Android accessibility has witnessed a dramatic boost from almost 0 access to picture descriptions to a wide range of options.
1. TalkBack 14.1 can describe images. Though, IMO, not as accurately as VoiceOver, it works well, and its auto-text extraction is awesome.
2. As you know, Seeing AI is now on Android along with its AI-oriented goodies.
3. @bemyeyes Be My AI just became available on Android. It's not yet capable of receiving pictures from other apps, but guess, hopefully, it will be added soon.
4. Since we're handling Google, I don't know when, but Lookout's AI capabilities, currently limited to users in the USA, will expand to other regions.

reshared this

Peter Vágner reshared this.

Let's Encrypt will issue new intermediate certs in Q1/2024: groups.google.com/a/mozilla.or…

Make sure your LE cert deployment logic includes serving the right intermediates that ACME should hand you, not just that same old LE intermediate you got years ago. Otherwise, there'll be breakage...

#x509 #pki #LetsEncrypt

reshared this

Peter Vágner reshared this.

The universe has cried and Kostyantin, the dev behind Whatsapp+ and Unigram+ NVDA addons, has delivered: Github-Release-Downloader, a NVDA addon that will happily download the latest release of any Github repo you throw at it. It's quite rudimentary at the moment i.e. we need to enter the URL manually, it can't handle multiple asset files etc. but I was just able to download some NVDA addons just fine. It's definitely a start.
dropbox.com/scl/fi/n7xn2fsv70c…
If you like this one and would like to support the dev, all the info you need is at t.me/unigramplus. Thanks, Kostya, and greetings to Ukraine! #NVDASR #Accessibility #Blind

Peter Vágner reshared this.

in reply to Bri😻

@Brynify With all of that being said, I do agree this isn't screen-reader-specific, and that as many people as possible should benefit. My answer to that would be to build it as a web service: let me open a GitHub repo URL, change the domain and nothing else (e.g. "github.com" to "rlshub.com"), and give me the info I need. Build screen reader add-ons to automate that step if it doesn't feel intuitive enough. @Piciok
in reply to Bri😻

@Brynify Okay, I put a rough version of this together. Change "github" to "gitrls" in any GitHub repo URL, and get either the single asset for the latest release, or a list of assets if there's more than one. Some URLs to try:

* gitrls.com/cartertemm/AI-conte…
* gitrls.com/digitalocean/doctl
* gitrls.com/Brynify/typing_sett…

It will silently trim any unrelated parts of the URL, so you could e.g. do this and it will work: gitrls.com/NVDARemote/NVDARemo… @Piciok

reshared this

in reply to James Scholes

@jscholes @Brynify I have just tried with both the NVDA and Lossless Cut repo at github.com/mifi/lossless-cut/ each of them having multiple assets to a release and I got a 404 error. Maybe I spelled something wrong but I don't think so as I replaced the "hub" part from my address bar. hmmm...
in reply to Paweł Masarczyk

@Brynify A 404 for the NVDA repo is currently expected, because the installer isn't hosted on GitHub as a release asset. GitHub's web UI always offers a zip and tarball of the source code under the "assets" button, but the GH API doesn't consider those to be assets in the same way. As far as hgitrls is concerned, an NVDA release has zero assets attached.

I am just testing @Jami with a #screenreader on both linux desktop and android for #accessibility.
Initial setup is accessible on both the platforms.
Listing conversations I can't really say as I only have single contact.
Audio calling is working fine. I am impressed that the call setup took just a moment. On android controls like microphone toggle, speaker / earpiece toggle and hangup button are working fine.
I am unable to find out in call controls with the keyboard on linux.
On both desktop and android I can write messages.
On android I can read messages, find and execute additional actions in the popup menu.
On the desktop I can't read incoming and outgoing messages with a screen reader. I haven't discovered on how to copy them.

In conclusion comparing this to the tox chat the Jami is more accessible with a screen reader. Perhaps I will be able to figure out how to handle the calls with a keyboard shortcuts however the fact message text is not readable with a screen reader on desktop linux and perhaps other platforms sounds dissapointing. The idea and decentralized nature of this communication app sounds really amazing.

reshared this

in reply to Peter Vágner

Exploring @Jami further I think I can post a few more observations.
The desktop app uses QT6 for the UI. Apart of some ounlabelled buttons such as Accept / Reject incoming call and some tab controls on the main window most of the UI elements are clearly labelled and accessible from the keyboard.
The desktop app has a keyboard shortcuts button on the main screen which opens a tabbed dialog with accessible lists of keyboard shortcuts. I haven't yet discovered if these can be tweaked but the default ones are working well for me. For calls it's ctrl+y for accepting and ctrl+d for rejecting / hanging up. Letter m alone can be used for muting / unmuting the microphone.
I need to find out if there is a way to create global shortcuts or do some actions using commandline switches. If either of this turned out to be possible it would perfectly integrate with the desktop.
Sending and receiving files is something I'm going to try next.
@Jami
Peter Vágner reshared this.

Testing #SeeingAI on #Android. This is *not* a walkthrough or tutorial, just a very quick, very dirty demo. No more than that, so speech is at the usual speed I have it, etc.

reshared this

Peter Vágner reshared this.

My partner Deedra reported an accessibility issue in the Monal XMPP client
recently, and their reply included the following help request:

> If you are aware of an open source enthusiast with a bit of experience
> in writing SwiftUI as well as accessibility, please let him/her/them
> know that we would love some help.

I don't know anything about this stuff, but I hope I can find people who do.
Any takers?
Probably you should reach out to the project directly via email, to
info at monal-im dot org.
Please boost far and wide.

Mentioning @xmpp as well, here.

reshared this

Priatelia, poznáte simplemobiletools.com ? Je to sada veľmi jednoduchých ale funkčných open-source aplikácií pre android. Za projektom stojí slovenský vývojár Tibor Kaputa, čo som doteraz veľmi rád všade zdôrazňoval, kde to bolo možné. Sám autor v komentári potvrdzuje, že napozadí prebieha predaj týchto aplikácií resp. nie ani tak aplikácií a značky simple mobile tools ako používateľskej základne v obchode google play spoločnosti ZipoApps. Už tu na fediverse som zhliadol vtipné komentáre, že v kontexte autorovho priezviska, týmto aplikáciám je kaput. github.com/SimpleMobileTools/G…
Peter Vágner reshared this.

#AndroidAppRain at apt.izzysoft.de/fdroid today with 3 updated and 2 added apps:

* Vault: password manager compatible with KeePass
* Round Sync - Rclone for Android: file manager using rclone for the heavy lifting

At F-Droid, 35 apps were updated.

Enjoy your #free #Android #apps with #FDroid and the #IzzySoftRepo :awesome:

Peter Vágner reshared this.

Peter Vágner reshared this.

ChatGPT can reveal its training data, that includes personal information.

not-just-memorization.github.i…

Here, the authors used a prompt to instruct ChatGPT to repeat a word forever, eventually resulting in different text that can be linked back to the source; examples include "company" and "poem".

No doubt this "attack" (in words of the authors) will soon be intercepted, but who knows what other formulation of prompt results in the same behaviour?

Analysis:
arxiv.org/abs/2311.17035

This entry was edited (1 year ago)

reshared this

Peter Vágner reshared this.

LLaVA-1.5 is an open-ish AI model which can provide image descriptions and allow follow-up interaction, akin to Be My AI. The best part is that you can run it locally on your computer if you have an appropriate GPU... or very, very slowly if you want to use your CPU. I thought it'd be cool to hook it up to #NVDASR so you can get image descriptions for the current navigator object and then ask follow-up questions. So, I wrote an NVDA add-on to do just that using llama.cpp. github.com/jcsteh/nvda-llamaCp…

reshared this

in reply to Jamie Teh

Size is weird, I just took something. But it happens the same if it says 1920x1080 slot 0 - image loaded [id: 10] resolution (38 x 22)
slot 0 is processing [task id: 4]
slot 0 : kv cache rm - [0, end)
slot 0 - encoding image [id: 10]
{"timestamp":1701472031,"level":"INFO","function":"log_server_request","line":2601,"message":"request","remote_addr":"127.0.0.1","remote_port":33884,"status":200,"method":"POST","path":"/completion","params":{}}
slot 0 released (3 tokens in cache)

Zaujímavý článok z histórie dopravy v čechách a na slovensku autobible.euro.cz/clanky/karos…

I have just found out up to date version of #rclone can be used on non-rooted #android devices thanks to the #round-sync app. github.com/newhinton/Round-Syn… cc @IzzyOnDroid ✅

Steffen reshared this.

in reply to Peter Vágner

Die App kann es gar nicht geben. Die kommt doch aus einem Phantom-Ort:

certificate DN: C=de, ST=NRW, L=Bielefeld, O=Nüsse IT, OU=dev, CN=Felix Nüsse

de.wikipedia.org/wiki/Bielefel…

Aber wenn Du darauf bestehst: warte einfach auf den nächsten #AndroidAppRain :awesome:

Danke für den Hinweis!

Peter Vágner reshared this.

I'm delighted to announce that #Libervia has received a new grant from #NLnet #NGI0 to develop an #email <=> #XMPP gateway.

This project will not only convert normal messages but also transform mailing lists into #pubsub based forums. It includes UI/UX enhancements on Libervia, end-to-end encryption, and advanced handling of attachments.

More at: nlnet.nl/project/EmailXMPPgate…

reshared this

Peter Vágner reshared this.

Making the switch from Google Podcasts? 🔄 Check out our latest blog post on migrating to AntennaPod! 🌐 Discover why it's the ideal choice for podcast lovers - simple, ad-free, and built on open standards. Say goodbye to tracking and hello to a great listening experience! #GooglePodcasts #AntennaPod #PodcastMigration
antennapod.org/blog/2023/11/go…
This entry was edited (1 year ago)

reshared this

Peter Vágner reshared this.

After two and a half years of rewrite, #Fractal 5 is finally out! Get the #GTK 4 #Rust #Matrix client from flathub.org/fr/apps/org.gnome.… and enjoy new features such as #EndToEndEncryption, location sharing, or multi-account with Single-Sign On 🚀

:boost_ok:

reshared this

Peter Vágner reshared this.

On our date night last night, Shae introduced me to a new game that any* two people online can play, and it turns out to be really fun!

You go to Wikipedia, and one of you (take turns doing this) names something off the top of their head that they're pretty sure a Wikipedia page exists for. (If you're not sure, you can use the search without submitting it just to see if a page turns up.)

After you both acknowledge that you understand the target, you simultaneously (we did a 3-2-1 on the voice chat) click "Random Page".

The goal is then to see who can reach the target page first ONLY BY CLICKING INTERNAL LINKS. No external links or typing allowed. We did eventually allow the back button so you could get yourself out of dead ends, but tried to use it in only that way.

It's surprisingly fun!

(* If you use a screen reader or other access technologies, or if, like me, you read VERY slowly, you might not be very competitive if your opponent does not also.)

reshared this

Peter Vágner reshared this.

VScan: Turn your smartphone into any accessibility aid you can imagine with GPT4 vision

Here comes my next idea & project. If you can think of a system & user prompt that would turn GPT 4 vision into an accessibility aid, i.e. by telling it what to look for in images and how to tell you the output, and then if you can simply do photos against these prompts using your smartphone, then you can basically turn your smartphone into a pretty wide range of accessibility tools (color detector, text reader, expiry date extractor, navigator, etc.).
I decided to try this in practice, and the results are pretty interesting! Well, you can try yourself:
github.com/RastislavKish/VScan

Note the app has been designed such that it can be easily used both for tools creation as well as standard image recognition you may be used to do with Be my AI or my Vision project, or you don't even need to be taking pictures at all, you can use the app to simply chat with GPT 4V (the model has the same textual capabilities as GPT4).
Though note there is currently no chat history review functionality nor conversation truncation when the 4k token limit of GPT 4V is exceeded, so the app is not optimized for this use-case.

There are few rough edges by now, probably the most annoying issue I'm facing on my device is that Talkback shows the braille keyboard in a reversed position i nthe session screen due to the display orientation, I need to figure out how to make CameraX adapt for the current device orientation.

But that's mostly a minor issue, the main functionality works as expected.
Any constructive thoughts and opinions on this project are very welcome, and, if you get to create some interesting accessibility tools you would like to share, I would love to hear about them!

Happy Visioning!

Peter Vágner reshared this.

What a discovery: LMI interface for changing bios settings from the userspace on modern Lenovo laptops is a thing for quite some time already. It's built into the linux kernel since 5.17. I have supported Lenovo laptop for a few months and I have only found about it today. It's nice for IT professionals but for blind users like me it's great too.
in reply to Wesley M

Yes, for me it is working instantaneously between the voice prompt. I was also surprised by the quick responses. I'm not sure if that is due to the holiday here in the States and the responses are quicker or if it’s because I wasn't asking any complicated request, but I will be playing a bit further with it today. Regardless, the new Voice feature for the ChatGPT is just simply wonderful for those like me taking advantage of the free tier.
Peter Vágner reshared this.

Ultimate Arch Linux Install [CZ/EN]: LUKS, BTRFS, Dual Boot & More - Step-by-Step Tutorial
#archlinux #guide #linux
youtu.be/CteatauMLz8
This entry was edited (1 year ago)

reshared this

Peter Vágner reshared this.

First we had Mac get a GPT 4 image describing utility, now it’s NVDA’s turn. github.com/cartertemm/AI-conte…

reshared this

in reply to Jamie Teh

@jcsteh From what I've seen, this is one of the most common questions on the OpenAI dev forums and other Q&A sites by far; people expect it to work like the web version of ChatGPT out of the box. So on one hand, statefulness would be a popular feature. On the other, it could decrease token usage and hence API revenue, or make token utilisation less predictable. There could also be questions about how long they store e.g. cached images within a session, although I think the data usage ship has well and truly sailed on this one. @pitermach
in reply to James Scholes

@jscholes @jcsteh All OpenAI APIs work this way, including Chat GPT. There's no state, you always send the entire conversation history. This is most likely what web Chat GPT (or the Chat GPT backend) does under the hood. Even if there was state, it would be an abstraction at best with little to no impact on token usage. You just have to feed the whole conversation history to the model every time to make it do anything useful, particularly in such a heavily multi-user setup. That's how transformers function.
Peter Vágner reshared this.

Did you know @kde had an #accessibility inspector? Currently it's hidden in the build folder of libqaccessibilityclient but not for long. We started the process of making it a standalone application and a first release is planned in 2-3 weeks!

apps.kde.org/accessibilityinsp…

#a11y #linuxDeskop

reshared this

Peter Vágner reshared this.

Quick Tip: Disabling Video Message Recording on WhatsApp and using only voice recording accessibleandroid.com/quick-ti… #Android #WhatsApp

Peter Vágner reshared this.

Peter Vágner reshared this.

Here's a goofy little project that @Mike and I did yesterday. In the course of conversation we stumbled on the idea of a monastic order of programmers singing Gregorian-style chants about programming practices. Mike encouraged me to write such a song with the help of ChatGPT and record it, then he'd produce it. Here's the result, a satirical ode to refactoring, clean code, and programming discipline in general. Lyrics to follow.

reshared this

Peter Vágner reshared this.

Na HitHitu rozjela kampan aplikace Librity. Neco jako cesky Audible.

Zajimave je, ze pri jejim predstavovani mel Matej na sobe mikinu @ceskodigital. Coz je cool.

Kouknete na to.

hithit.com/cs/project/12353/ct…

#knihy #hithit

Peter Vágner reshared this.

Peter Vágner reshared this.

Vision: Access GPT 4 visual capabilities from your laptop

ever since Be my AI demonstrated the capabilities of GPT 4 for understanding the content of images, which is not a single bit less impressive than its textual abilities, I was looking forward to having an API I could use.
Because this brings awesome potential for various use-cases, though also for economical reasons, Be my AI is not on Android yet and I can't really afford spending $20 monthly on a ChatGPT + subscription.

Few days ago, OpenAI finally introduced a vision API, and here is my really simple, but working Be my AI in terminal:
github.com/RastislavKish/visio…

Let's have some fun, there is a lot to enjoy!

Note: Uploading images can take a while, though if it apparently takes too long, it's likely frozen, OpenAI is dealing with a lot of demand after their dev conference on six-th November, so chatGPT is down from time to time. My script is not very well prepared for this at this moment, I may add proper timeouts and retry options if it becomes too big of an issue, though I rather expect things to calm down again as things get resolved on the OpenAI side.

Peter Vágner reshared this.