bscross32

I've been told by multiple people many times that spinning a dryer drum by hand is gonna fuck up the belt. I call bullshit. I've been doing it my entire life and it's never once caused a problem. I think it can if you're a dumbass about it and spin it really fast over and over. Indeed, you can actually feel the belt start to slip if you spin it too fast, so then back off a bit.

Why do I do this? Simple. I spin it just enough to where there's a gap in the fins at the bottom, then I sort of funnel all the clothes into a pile and push them from back to front and pull them out. I find it easier to manage that way than if one of the fins is at the bottom which basically splits your clothes into two piles.

I think the people who do this are the same type of people who get in their car when there's a lot of snow on the ground and just fucking gun it thinking they can get unstuck that way. No nuance, no feeling. I feel like I'd have a better chance getting a car out of a snow drift than some people out there.
#blind #appliances #bored #Friday

New version of #Pubsplash released. Firming up some API stuff including ElevenLabs. No more need to fetch voices manually, it's done for you and results are cached and compared against the cache at startup. Put in your keys and hit validate for any that haven't been already saved.
More parameters for API-based tts services have been exposed
Fixed a nasty crash when deleting plugins from running sources
Note: If you ahve Waves plugins, you will currently only see one, this is a known issue and may be addressed in a future version
Each tts engine retains its own settings and they'll be recalled when you switch back to that engine
Optional streaming (on by default) for ElevenLabs V2 to reduce latency
Selecting models now uses a sdropdown list instead of a type in field to make it easier to get what you want
There's an API tab that will help you track your usage per session. Only ElevenLabs allows checking your credit balance, and it's a refresh pres away now
You should no longer have to monitor tts sources to hear their speech.
tts sources set to ElevenLabs now show the name of the voice rather than the voice ID
ElevenLabs speech rate can now be utilized without error
Your Audiopub password will no longer be stored in plain text in the configuration file. The change over should be seamless.
If Audiopub crashes, a crash dump will now be generated and stored in a crashes folder alongside your configuration file. If this happens, file an issue via GitHub and attach your logs and the dump

https://github.com/ironcross32/pubsplash/releases/tag/v0.1.3

0xc0000005, STATUS_ACCESS_VIOLATION
These are fun...

Pubsplash can now load VST2 plugins. VST3 will come later. I'm choosing specifically not to handle 32-bit plugins. This is the quickest method of streaming to Audio Pub I know of. The first time you ever use it, you go to File > Configure audio pub, enter your credentials and hit connect. It will connect on startup. There's always a default scene that can't be removed with the bare minimum of your microphone and a tts source for reading chats (both can be removed). So if you hit edit on those and they're using the right stuff, you're good. The workflow to start streaming is:

1. Launch the app
2. Hit ALT+S
3. Enter your stream title, description, and check the box if you want it archived
4. Hit OK

That's it. Everything as it should be, you're streaming. If you have to stop, hit ALT+T. That information is kept for you per app session so if you want to start again, ALT+S and it just happens.

I'll use #pubsplash for posts relating to this app.

OK, this thing BSI does where you realize you made a mistake, back up, fix it, but the letter you added is a single letter contraction which will get translated when you hit space even though there are letters before it which makes that rule invalid is pretty fucking stupid. Fix it... like, now.

If you have an air purifier, today might be a good day to use it. Also be glad it's not an air pubifier, because that would just be unpleasant.

Sometimes, these two commands can alleviate weird network issues on a single Windows machine, especially if the installation is old. The latter needs elevation, so run from an admin prompt:
ipconfig /flushdns
netsh winsock reset

If you own a Lenovo PC, you may have noticed that adjustments to machine-specific functions are spoken via SAPI when Narrator is enabled. This includes things like the camera being on or off, screen brightness, touchpad status, power profile, etc.
Of course, it'd be great if this were screen reader agnostic, but it relies on narrator.exe being running.
This little program will trick the utility into dispatching the messages, because it runs as narrator.exe, even though all it does is creates a system tray icon with a context menu. Run it normally or place it in shell:startup to have it run when the machine boots.
It's worth noting that Narrator won't start with this being active, so you'll need to exit it first if you wanna use Narrator.
https://www.dropbox.com/scl/fi/ea9onvco20e4jb0u6vy26/narrator.exe?rlkey=3b93hr84q2k9h4il7k1wiknd2&dl=1

I've been using Firefox a bit of late after not having done so for years. There are some nice accessibility things, but one aspect that irks me is that every time I launch it, it sends a toast notif asking if I'd like to set it as default. If I hit no thanks, it pops up the settings dialog in Windows anyway. It will also continue to ask every time it's launched.

Claude is coding me an audio FPS style game in Rust using Bevy and Steam Audio. I never thought I'd see the day #blind #gaming #a11y #AI

Thanks #Suno for making checkboxes appear as buttons to screen readers in your web UI #blind #accessibility #fail #AI

I'm seeing an ever-increasing number of people investigating AI solutions that operate your computer for you. I must admit that I have great trepidation towards this notion. Of course, I can't look past the obvious benefits of being able to operate inaccessible software.

I tried Guide AI a while back, and it simply didn't meet my needs. I wanted to use it in full screen applications such as games, but you had to type to it, meaning, you had to leave the window. While it did offer a keystroke to jump quickly to its own window, it wasn't a good enough solution in my opinion.

Apart from the typing limitation, I found that it couldn't even perform the most simple tasks reliably. I didn't see value in it, because I kept running into message limitations in the trial. The model would take circuitous routes to do very basic things, such as opening task manager to verify that an app was running. It also launched the Xbox app in a way that left it stuck on the splash screen. Now, I don't have very many complimentary things to say about the Xbox app, but it at least has never done that when I launched it myself.

Naturally, this put me off using the service, and I canceled the trial, knowing that it would likely improve and that I'd be stuck paying more if I ever wanted it. The one good thing it had going for it though, was a kill switch. It narrated every action before attempting to perform it. If it was about to do something you didn't approve of, you could hit that keystroke and the entire task chain would be aborted.

Many of us are using Gemini, Codex, Claude code, and other agents. Those of us who've been doing this long enough will know that while the models can produce excellent results, they can just as quickly go down a path full of strange actions. They will proceed, full steam ahead, with great aplomb, down the wrong path. If left unsupervised, the best case is that you've burnt some tokens and you'll have to burn some more to have it walk back the changes. I don't think I have to enumerate what the worst case scenarios could be.

All this to say that I think I'll keep operating my computer without any assistance from AI tools for the time being. It means I'll miss out on some potential benefits, but I'm all right with that, and feel more secure in knowing that an AI agent isn't about to run amuck on my system. There are those who will read this and not like my position, which is fine. I've seen a lot of people snap to judgement calls and use terms like, "Ai hater", which honestly leads me to wonder if continuous use of LLMs have whittled away at their critical thinking skills.

With technology like this, one should always make an informed decision. Jumping on board with every single development in the field of generative AI is a choice, but it's one fraught with consequences.

Here's me typing on Mac. Tip tip tip tip chigga tip tip tip chigga chigga tip tip dup tip tada tada tip chigga woop tip tip tip woop tip chigga dup dup dup chigga woop woop woop tip tip tip tidda tidda woop woop woop.

Autocorrect, spelling, all that shiiiiiiiet needs to go, now. And... done.

The older I get, the more my brain resembles an LLM where the context window is rapidly approaching and information needs to be summarized before exceeding it.

It kinda bothered me how the Unspoken NG addon for the 64-bit version of NVDA alpha/beta cut off reverb tails every time a new sound played. I'm very snobby about this, I don't like cut-off reverb tails anywhere lol, so now it's got a reverb bus that continuously feeds nvwave.WavePlayer frames of tail data up to a configurable limit. You also have the option of decreasing or increasing frame size at the cost of added latency for systems that can't handle the default.

If you acquire a new Windows laptop and would like to know what the Fn+F1-F12 shortcut keys do, try pressing them while Narrator is active. This worked in my case. The speech was much slower, and picked Microsoft David even though my selected voice differed, but it did read out things like touchpad enabled / disabled, brightness X%, and even Camera function disabled/enabled for a physical switch on the right side of the device. NVDA did not announce any of this. No guarantees it'll work for your machine, but worth a shot. #blind #accessibility