Notices by Simon Jaeger (simon@procrastodon.net)

With a bit of input from Gemini about its own API, I wrote two shell scripts to:
1. Provide a text-only alternative to an audio or video file.
2. Create a timestamped description of the visuals in a video file, not including audio-only information.

Gemini definitely has some flaws when working with video, and I've only tried it on a few files. Interestingly, it gained enough context clues from a one-minute clip of The Big Bang Theory to successfully identify a character who is never named in that clip. It also doesn't know how to hold off on giving names to characters within its descriptions until those characters are given a name within the clip. If I give it permission to use context clues from the audio, it uses later ones to describe earlier parts of the file.
And it's not perfect at separating audio from video--for instance, my visual-only description had sentences like, "Sheldon walks into the kitchen and makes a request." and "Sheldon begins to sing." I could always mute the audio when sending the video, but I was hoping I could get Gemini to understand the difference and produce reasonably useful video descriptions.
I'm mostly using 2.0 Flash because I *think* it's still free for now, but I'll compare 1.5 Pro at some point and keep tweaking the prompts.
For audio, it's great. I tell it to add paragraphs for readability and line breaks to singing, and it does this quite well. It can't always identify sounds, but it tries pretty hard.

For some reason, in Thunderbird 128, NVDA is no longer reading the bottom list item, or counting it as part of the list. So whenever a new email arrives, I just have to open it and see what it is. Anyone else experiencing this? Do I have to fake an email from the future now?

@fastfinge Are we really marking anything remotely related to harry Potter as transphobic? That feels like going a bit far. I welcome disagreement from someone who has more skin in the game but I feel like it promotes the idea that supporting anything related to Harry Potter is also supporting the author as a person. Nothing about this post is transphobic, unless the fanfic is, which is something we'd probably want to know.

Apple seems to have discontinued the lightning to headphone adapter. They're still selling the iPhone SE and 14 on the online store, but if you want to use headphones with them, I guess you're shit out of luck. Every day, I enjoy being an Apple user a bit less.

@greengaybles Go to a URL like this one, and if you're on the phone, patiently read through it. https://kind.social/api/v1/instance

I feel like all I can do is laugh at the last week of insanity.
I had a few work demos due at the end of the week. I've already been putting these off because recording demos caused me an unreasonable amount of stress due to having no easy way to edit them and thinking myself into a hole. Thanks to mastodon, I discovered LosslessCut, and I have to first say that this little tool has been amazing and has saved me immeasurable time and stress. I can finally do basic edits to my videos instead of just sending in a fumbling mess.
So I got a few demos done on my iPhone and was able to cut out the fumbling bits and submit them. Great!
Then, as I was gearing up to record the remaining demos on Wednesday, my mother's car broke down and she needed to come to my apartment so she could stay in a place where you can actually function without a car. That's my private recording space gone.
I should have had them done on Friday. But my mother decided to stay until Saturday. Then Sunday. I got plenty of recording done, but it just wasn't enough.
Today she left, and I tried to do the final recordings on Windows.
Game Bar (which is the default screen recorder for Windows) absolutely refused to work on either of my machines. On my Windows 10 laptop it recorded incredibly choppy audio. On my Windows 11 Surface, it gave me a cryptic error.
So I thought, "Why don't I just join Zoom and record myself? That should work, right?"
It did work, but it kept popping up alerts telling me I was muted, and those would steal keyboard focus. Because I wasn't recording my own microphone audio, I didn't want to be unmuted, so I just dealt with the obnoxious popups and edited them out with LosslessCut.
Then I did a few demos where I actually did need to be unmuted, and after hunting for the "Original Sound" option, I realized I couldn't turn it on while sharing my screen, so had to un-share it, turn it on, and re-share. Annoying, but fine.
Finally, I got done with all the rest of my Windows demos, looked at the file size, and realized they were all under 10 MB.
So I called Aira. Sure enough, blank video. Absolutely nothing there. MP4 files with only audio on them.
At this point, I just want to collapse on the floor, but I did something I should've done years ago and installed OBS. It took me 5 minutes to figure out how to make it record my microphone and system audio simultaneously, another 5 minutes to set up some hotkeys, and then I was finally, finally able to record my demos. Bonus: They were in MKV, so even though they do have video, they're still tiny.
I am so many kinds of done with today. Which is timely, because it's tomorrow.

Happy fucking Monday, world.

@Fragglemuppet I'd say go into ctrl+nvda+v, shift+tab past the OK button, and you'll land on the modes selector. uncheck "Off" and "Beeps", so your only modes are "On" and "On-demand". Now if you press nvda+s, you'll change speech so it responds to keyboard commands like nvda+t and nvda+f12, but doesn't randomly talk unless you specifically request information.

@Fragglemuppet Hmm. That makes me wonder if some other window is popping up just long enough to affect NVDA's focus. Do you ever read books or long webpages on the computer, and does it happen there too?

@Fragglemuppet Does this only happen in a particular app or a particular type of control? I don't usually have this problem. I also usually turn speech off when i'm not using it. You can remove the speech modes you don't use now, so it's just an on/off toggle instead of cycling through all the modes.

You're a taxi driver. You're sent to pick up someone named Simon. You drive up and there's a person with a cane standing on the curb, apparently doing nothing and facing toward the parking lot. Do you:
1. Call out to the person.
2. Honk at the person.
3. Sit in your car doing nothing until a random stranger comes up to the blind person and tells him there's a taxi 3 car-lengths away, sitting there, doing nothing.
I feel sometimes like all of humanity is being enshittified.

@Fragglemuppet @Bruce @techsinger @FreakyFwoof In a technical sense it's better because anyone can have their own spot in the fediverse, and either invite people or keep it to themselves.
In a cultural sense it seems better because people actually recognize the existence of others unlike themselves, whether that's people who might appreciate content warnings or alt text.
The user experience is very Twitter-like, and the blind community basically migrated off of one and onto the other, so that's going to feel very similar. But I used to look at mainstream Twitter sometimes and while there was good content there too, it was a lot more full of low-effort garbage. If I tried to advocate for alt text, I'd get replies from people who either didn't understand why I was asking for it (even though I specified in the post) or didn't care. If I replied to people asking questions about blindness, I'd get shorthand disbelief. Twitter used to only allow 140 characters and I think it trained people to stop caring about what someone had to say after the 141th character.
So I feel like it's more accurate to say Mastodon is like the best of Twitter. And a lot of people who were in specific interest-based bubbles or the disability space didn't experience the worst Twitter had to offer.

@Fragglemuppet @FreakyFwoof ... You could have an account on every individual forum and that was fine. Now that's still fine, but all the forums talk to each other.

@Fragglemuppet Are you unhappy with your current one, or just wanting to try something else?

Tried to activate JAWS. JAWS said "you need to restart your computer to be in demo mode otherwise I'll close and not let you activate me." Fine. Restarted computer. Computer wasn't back after 5 minutes. Turns out it's complaining about the power output of my Dell dock being 60W instead of 65W. Seeing AI can never read the button that dismisses this and I always forget. It's escape, so I shouldn't forget. Laptop boots, but it needs to update. I leave it for a while. Come back. No sound. It's rebooted again and I have to dismiss the error again. It boots up. Sound disappears but I can read the login screen with OCR. After a few minutes of frustration, I remote to it from another machine. Turns out my Aftershokz headset had turned on in the bedroom and my system audio switched to it. Now I'm ready to activate JAWS. It's not even Thursday.

I think I would literally pay $1,000 for an app that sits in the background and records microphone, system audio, and system video, then gives me trim/cut capabilities along the lines of GoldWave. I'm always being asked to record demos for work, and I apparently do them really well because they keep coming back for more, but I spend 90% of my time avoiding it, 9% of the time obsessively nitpicking and planning every detail, and 1% of the time actually recording the demo.
I tried to edit existing video in Reaper, and it works but produces weird, inconsistent results. If I try to actually save it as MP4, I get really low colour" depth and other issues. I just want to open up a video file and cut out all the extra junk and failed sentences. It should not be this hard.

It would also probably be easier if I could make a script and then reading from it, but I need hands to read and hands to use my devices. So I try to rehearse the script, but end up getting it wrong and recording 500 failed takes because I focus too much on saying the right thing. So then I improvise, but I end up recording successive takes that get longer and longer as I perfect the whole thing.

@Fragglemuppet Wow, this is very broken for me. As soon as I paste your ID and press next, it just sits there loading infinitely. VoiceOver can't read a single thing.

@Fragglemuppet Strange, I've got nothing. Can you send me yours? Perhaps if we both add each other it'll sync up. Annoyingly, I now remember having this problem when I tried it a couple of years ago, so it also might be me.

@Fragglemuppet I got that a long time ago, never got to test it with anyone. 05b73b7e226da819b9faed41bca45ed55c3c587b90bc676f0ba64c0874909bba14

The OnePlus setup wizard has a little intro screen that says "Welcome. Are you ready?" Followed by a next button which is completely invisible to Talkback. Notably, ChatGPT helped me find it.
Feels like a very nice phone so far. Not a great start on software, but I can't say I wasn't warned.
We'll see how the rest of this goes. I'm keeping all the packaging.