simon

@simon@dragonscave.space

Accessibility tester, bookworm, occasional musician and audio producer, coffee addict, selfhoster, tech tinkerer, blind from birth.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

simon, to random

I'm finding that if you use the latest fine-tuned Demucs model (htdemucs_ft) on a true lossless song, the result is very, very close to perfect separation. Sometimes I actually can't tell the track was separated using AI. The jump from MDX to this new model seems as big as the jump from Spleeter to Demucs. The problem is that finding lossless songs is hard if you don't know what you're doing, and when you try with a lossy song, you quickly become aware of just how much data the compression removes. Even some of the lossless tracks I get from Deezer are not truly lossless, which is annoying. I don't have an accessible way of verifying it, so I could easily have hundreds of GB of fake FLAC songs. But meanwhile, the results I get are mind-blowing.

simon, to random

Yikes. iOS 17 betas include a way to intentionally disable all sound when VoiceOver is loaded, including speech. It's called Sound Curtain. On my phone, there's no gesture for toggling it, it's just an option in Quick Settings and in Voiceover's audio settings. And it persists across Voiceover restarts. If you accidentally turn this on and then lose your place on the screen, you're screwed unless you have a Braille display. I'm not sure what I think of this; I'd put money on at least one person digging through settings and accidentally turning it on with no way to undo it. And it's not like there's a set of instructions another blind person could pass along for fixing it. "Open quick settings, keep flicking right and double-tapping until your sound comes back?" Yeah, no. For now, I assigned four-finger quadruple-tap to toggle it on and off. Hard to do by accident, easy to remember.

simon, to random

I wonder why everyone in blindland seems to be getting the 15 Pro Max this year. Or rather, anyone who is getting a new phone seems to be getting the Pro Max. I did because people on Reddit said it felt lighter and nicer in their hands than the previous pro max phones. They were right; I've had it for half a day and it doesn't feel uncomfortable at all. I mostly forget I'm holding a new phone until I think about it.

simon, to accessibility

/ awareness in a nutshell

simon, to random

I bought the Goyo replacement (Supertone Clear). Two things I hate already:

  1. It only works on two devices at a time. I don't know what the implications of this are if you lose one of the devices and can't unregister the plugin, or you want to run it in two diferent DAWs.
  2. In order to download it, I had to download the trial. To download the trial, you need to enter your e-mail address, accept the data usage policy, and accept marketting emails. So I had to sign up to their newsletter in order to download the thing I bought. I'm already signed up, but that's not the point.
    Just ... sigh.
    By the way, the super secret promo code given to beta users so they can buy the plugin for $29 is BETA29. But shhh, it's a secret.
simon, (edited ) to random

I stretched AI audio tools to their limits today.
In 2011, Taylor Swift did a live cover of Train's "Drops of Jupiter" (Tell Me) during her Speak Now world tour. It was never recorded in studio. So I decided to see whether I could turn a live recording of a guitar cover from a crowded stadium full of screaming fans into a clean and listenable studio vocal that could be mixed over the original instrumental. This required heavy use of Demucs and Goyo, and I also had to clean up the ambiance track, stick an aggressive high-pass filter on it, and mix it back in so I could resurrect some of the sibilants. And being a live guitar cover, it wasn't done at any particular BPM, just whichever one sounded good--which actually turned out to be quite a few, ranging from about 150 to 166. If I finish this, I'd rather find a good guitar instrumental, but for now, here's a 20-second sample of the source material plus the first verse and chorus. I think this is a pretty crazy example of what AI can do for audio in 2023. I can only really take credit for adding the stretch markers and doing some very basic mixing.

simon, (edited ) to random

I don't know if I could ever realistically not be a Windows user, but i'm so tired of its "I know better than the user" mentality. Windows home blocking you from creating a local account. All versions of Windows force-rebooting for updates even when there are active applications that don't want to be closed. Windows 10 badgering the user to upgrade to 11 and only providing a "Remind me later" or "No, not right now" button. Constant setup screens that block the desktop from appearing (and block me from opening my screen reader if it's not set to start up the right way), and have nothing but a "Remind me in Three Days" button. The operating system dictates the questions and answers, and the questions are bordering on aggressive while the answers are all polite no's or enthusiastic yesses.
I don't game very much, and have always separated my casual use and audio production hardware. If Linux accessibility were better, I would probably be a Linux user.

simon, to random

I don't think VIPMud was a net positive for the blind community. It's terribly slow and hasn't had meaningful updates in years, not to mention it costs money, but somehow people are still holding onto it. MUSHClient has a more confusing interface, but under the hood it's a way better client, and writing code for it is even easier. But soundpack developers don't want to learn to code for MUSH, and players flock to whatever client has the best pack, so we end up with incredibly good soundpacks made for an incredibly bad client. If VIP had just not existed, we wouldn't have this split, and someone would probably have developed something modern and cross-platform by now. Mudlet might end up being that something. Maybe the only way to get users away from old software is to make a soundpack converter.

simon, to random

Every once in a while I manually zip an audiobook with more than 100 files, which reminds me that my automatic book packing script has code to rename files 1 through 99 to have a leading 0. Why, you ask? Because VoiceDream Reader has literal file sorting, so it plays files 1 through 10, then 100 through 109, then 11, then 110 through 119, then 12 ... and so on.
In light of the subscription kerfuffle and complete lack of transparency from ... whoever owns this thing now, I should probably start using Book Player for audio and really test out Speech Central so I can send in some feedback for it. This is one of a growing list of really stupid VoiceDream bugs that I just automatically work around because they've been there for such a long time.

simon, to random

VoiceDream change log: Bug fixes, bug fixes, bug fixes, bug fixes, bug fixes.
Me: Hi, this app crashes when I seek through a book quickly and I have to delete everything twice.
Change log: Bug fixes, bug fixes.
Me: Is this a feature then?😕

simon, to random

For years, i've wanted some kind of virtual reality voice chat application. At a party or other social gathering, you can wander around and talk to people while hearing parts of the conversations around you. Nothing available to us really simulates that in a good way. Being able to walk around a virtual space with binaural audio and proximity effects would be really interesting. Even a sidescroller situation would be interesting, and maybe ultimately less confusing. The demand for it probably isn't there, and I don't do much group conversation lately, but it's still an idea I find appealing for some reason.

simon, to random

Q: Is there any purpose at all to the Be My Eyes waiting list? Because it seems like lots of people are being given priority access and 0 of the people who signed up for it on day 1 actually have access to it now. Should I just roll my own solution to send an image to GPT 4 and ask a question about it? (I don't even know if I can do that with the API I have, that's how not in the loop I am apparently.)

simon, (edited ) to random

Here's a recording from a very noisy microphone setup which Goyo cleaned up shockingly well.

simon, to random

Just updated an LG Velvet to Android 13 (took them long enough). Eloquence stopped working, and the play store is claiming that Vocalizer is unavailable on my version of Android.
I have a Pixel that uses Vocalizer and an S20 that uses Eloquence. Both are on Android 13.
Something doesn't add up here.

simon, to random

Does anyone else worry that Netflix is looking at the numbers for people watching audio described content and going "Holy shit, why are we doing this when nobody is using it?"
And really, it's just because people pay for Netflix to support the content and then play the content off audiovault so they can background it.
Is this less common than I think it is?

simon, to random

To those who are in search of a good reading app for iOS that isn't VoiceDream and are willing to provide good feedback, I'm in touch with Dolphin regarding Easy Reader, and they seem interested in feedback, both from me and from others. You can try the app for free and report any findings or suggestions to support@yourdolphin.com. I've reported an unnatural pause when reading with the Eloquence voices. (You can change this from voice settings, but even when it's set to 0% there is still a slight pause after paragraphs.) In many ways it seems to behave similarly to VoiceDream reader, and it is significantly more stable in my experience. It doesn't have the nice variety of book sources so you have to share files out to it, and that's one thing I may suggest to them, but once you do add a book, reading controls work well and the headset can be used to control the book, much like VoiceDream.
I will probably do the same experimentation with Speech Central, but I hear a lot more people talking about that app, so wanted to look at something that's been around for some time but isn't really being looked at.
Easy Reader also does audio the same way VoiceDream does, including support for some DAISY features that aren't in VoiceDream, such as skippable structures.
I don't have a lot of experience with Dolphin, but I have gotten very good responses from them so far, so am hoping their developer(s) are just as receptive to feedback.
There is also an Android and Windows version. They all make you log in with a Dolphin account, but there doesn't seem to be any kind of book synchronization, which is unfortunate. If i'm wrong, i'll edit this post.
It's a work in progress and I think it can only get better if more people use it.

simon, to random

On an only slightly related note, what are people using for a screen reader friendly Mastodon experience on Android? I'd love to see something that implements Talkback actions properly.

simon, to random
simon, to random

Does anyone know where Apple has hidden the “Open VoiceOver at the Login Screen” setting on Ventura? Can’t find it in VoiceOver settings, Users & Groups settings, or VoiceOver utility.

simon, to random

As my 500GB SD card ful of books slowly fails and I'm left with nothing but a USB C Samsung T5 for portable data storage, I really feel the lack of USB C on my iPhone. To get files off this drive, I have to carry around a camera kit (which is already larger than the SD card reader), along with a USB A to C converter, because there is no camera kit that directly supports USB C for some reason. I guess if they made that, it would take away from the value of an iPhone 15. Will I get an iPhone 15 just to make this problem go away, even though my 13 works fine for every other use case? Time will tell, I guess.

simon, to random

I am about to cancel my Dropbox subscription, which involves clearing out a ton of huge shared folders from my Dropbox. So I'm going to put this out into the world:
If you are the owner of a shared Dropbox folder, I would really urge you to consider sharing it elsewhere, either in addition to or instead of Dropbox. OneDrive, Google Drive, and other storage services allow free users to join a shared folder without filling up their storage. Dropbox does not, because they want people to pay. So many people are paying for Dropbox just for the privilege of having a copy of someone else's folder on their own account. The last straw for me was when I noticed that my family subscription--which costs me over $200/year--was full because I shared a large folder with someone else in the family. Dropbox doesn't store the folder twice, but they count it toward my family storage twice. Meanwhile, every time I log into the website I get a popup telling me I should upgrade to Dropbox for business, for even more money. So I'm pretty much done here. You can try this at home.

simon, to random

I'm now using Goyo as a rule whenever I'm on TeamTalk with someone. I never realized how mentally taxing background noise could be. Goyo doesn't eliminate it unless I turn the relevant slider all the way down, but it does reduce it and make the voice stand out more.

simon, to random

I ran a recording of some kittens talking to their mother through Whisper. The result was in another language, so I ran it through Google Translate. What I got in response was mildly creepy. Here it is in full:

Again? It's you. It's you again. What are you going to do? You?

And here's the original text from Whisper:

またか? あなたがよね 再び あなたがよね どうします? あなたがえっ

simon, to random

I could've spent a whole bunch of dollars on a proprietary book reader from a profit-driven assistive tech company, but instead I spent a fraction of that money on the impossibly tiny Unihertz Jelly Star, an Android 13 phone with a 3-inch screen, dual-SIM, micro SD, 256 GB of storage, 8 GB RAM, and (are you sitting down for this?) an off-screen fingerprint reader and a headphone jack. It's very slow—my Galaxy S10E from 2019 beats it slightly in both single and multicore, but it's plenty usable and absolutely delightful. I've wanted a Unihertz phone for years, but by the time I heard about them, their phones were all running outdated versions of Android or they had other compromises. I have Eloquence working on this, I can put a whole TB of storage in if USB C + 256GB internal is somehow not enough, and it was less than half the price of the base iPhone SE. It feels like a toy, and maybe it is, but it'll be a really fun one.

simon, (edited ) to random

: Have you ever broken something made of glass?

  • All
  • Subscribed
  • Moderated
  • Favorites
  • anitta
  • InstantRegret
  • mdbf
  • ngwrru68w68
  • magazineikmin
  • thenastyranch
  • rosin
  • khanakhh
  • osvaldo12
  • Youngstown
  • slotface
  • Durango
  • kavyap
  • DreamBathrooms
  • JUstTest
  • tacticalgear
  • ethstaker
  • provamag3
  • cisconetworking
  • tester
  • GTA5RPClips
  • cubers
  • everett
  • modclub
  • megavids
  • normalnudes
  • Leos
  • lostlight
  • All magazines