Anyone here a fan of NPR's Wait Wait... Don't Tell Me!? This week's show is hosted by Iranian-American comedian Negin Farsad. I'm a fan of both Wait Wait... Don't Tell Me! and Farsad, so I'm really looking forward to this one! Peter Sagal, the show's regular host, isn't hosting this week.
The ChatGPT website has undergone some changes, and since yesterday, I've no longer been able to press Enter on the edit box to get NVDA out of Browse Mode so I can start typing. Instead, I have to focus the edit box and press NVDA+Space. Interestingly, Enter works if there's already at least one character typed into the edit box. But when you want to type something for the first time, or start writing a new message later in an ongoing conversation, it doesn't work. This happens regardless of whether I'm using an NVDA alpha, beta, or stable release. JAWS, however, doesn't have this problem.
@NVAccess
RE: https://dragonscave.space/@amir/117409083748249946
Update 1: openevv now has proper pause settings, Always/ At the end of sentence/ Disabled, to match Eloquence. The NVDA add-on also supports pronunciation dictionaries, but the dictionary files currently generate delays upon switching to other engines, or may silence NVDA as they are edited or saved.
@Mudb0y
The Microsoft Neural Voices Manager and Microsoft Neural Voices Driver add-ons for NVDA have been updated. If you use these add-ons to enjoy high-quality offline neural voices, this is definitely an update worth getting. And yes, for the first time, I also have two offline Neural Persian voices!
Here are some highlights from the changelog:
• Added support for new languages, including Amharic, Azerbaijani, Bengali, Bosnian, Persian, Filipino, Gujarati, Armenian, Indonesian, Javanese, Georgian, Kazakh, Khmer, Kannada, Lao, Macedonian, Malayalam, Mongolian, Marathi, Malay, Maltese, Burmese, Nepali, Pashto, Sinhala, Somali, Albanian, Sundanese, Swahili, Tamil, Uzbek, Zulu, and others.
• Added support for new countries and regions corresponding to the newly supported languages.
• Added 33 new languages, bringing the total to 77, and 25 new countries, bringing the total to 105.
• Added 94 new voices, bringing the total to 349 available voices.
Download links - pages are in Russian and Google Translate with Chrome makes it easier:
NeuralVoiceManager:
https://nvda-addons.ru/page.php?id=NeuralVoiceManager
NeuralVoices Driver:
https://nvda-addons.ru/page.php?id=NeuralVoices
openevv progress report: bringing it closer to Eloquence:
1. Two punctuation-related issues that affect ViaVoice but aren't issues in Eloquence have now been fixed in openevv. They were also in my report:
• Opening parenthesis pronunciation when followed by another symbol.
• Period, colon or exclamation mark following a closing quotation mark.
My tests show that the other ViaVoice-related glitches are still present, but good progress is being made toward resolving them.
2. The SAPI 4 phrase prediction style issue has been fixed.
The openevv NVDA add-on now includes a checkbox to disable phrase prediction. Phrase prediction is disabled by default.
3. The severe sluggishness when navigating environments such as Fedra messages and FastSMRW messages has been addressed.
Eloquence 64 and 64RS are still slightly more responsive, but the gap has narrowed considerably, and openevv's responsiveness is now at an acceptable level.
4. openevv still has no features or settings for pause reduction, and its pauses are long.
This is still annoying for those of us who are used to reduced pauses in NVDA, as well as for JAWS users coming from Eloquence, which has minimal pauses by default.
5. Finally, the higher sample rate issue remains unresolved.
The way openevv handles higher sample rates, including its checkbox for "Wideband above 11 kHz," is far from satisfactory. Either the options don't work at all, or enabling that checkbox makes the audio quality downright awful at higher sample rates. A better solution for higher sample rates is clearly needed.
Overall, progress is good, and I want to thank @Mudb0y for staying on top of these issues and continuing to work on them.
@nick @fastfinge
Fedra version 0.7.1 is out, containing mostly bugfixes and a couple additions. Changelog:
- Added `Home` and `End` to the media player, to go to the start or end of the media.
- Added an option to start Fedra maximized.
- Added options to turn off streaming, and to set how often Fedra checks for new posts when it isn't streaming.
- Fixed closed dialogs and media windows staying in memory and leaving Fedra running.
- Fixed Fedra crashing on startup on Windows when PC-Talker isn't installed.
- Fixed Fedra creating a lock file in your home folder on macOS.
- Fixed VoiceOver announcing menus when you press shortcuts on macOS.
- Fixed filters and muted hashtags not working in remote instance timelines.
- Open User Timeline and View Author Profile now let you pick anyone in a post, including the authors of quoted posts inside boosts.
- Turning notifications on or off for a timeline now says the name of the timeline.
Download: https://github.com/trypsynth/fedra/releases/tag/v0.7.1
Enjoy!
A few interesting IBM ViaVoice factoids!
Apparently, IBM really liked Samsung. When they got their hands on Eloquence and started tweaking pronunciations for what we now know as ViaVoice, they fixed the pronunciation of "Samsung". The "sung" part is pronounced correctly in ViaVoice.
IBM's engineers also changed the pronunciation of "identification", but they didn't exactly do a brilliant job with it. They messed up the "fi" part. They also seem to have missed the fact that if you're going to fix "identification", you might want to fix "identify" too. Perhaps someone forgot to put that on the list.
IBM also corrected "nationale" and "Italia". But there was a small problem: fixing "nationale" also breaks "internationale", so apparently the family tree of a word matters too.
And then there are words like "chased", "chasing", "unread", "neuroscientific", and others whose pronunciation IBM somehow managed to make worse. Why? I honestly have no idea.
There is, however, one IBM pronunciation correction that I can definitely get behind. For some reason I still don't understand, they also fixed my name. ViaVoice pronounces "Amir" correctly. Of course, "Amir" is also an English dictionary word meaning "prince", so this has absolutely nothing to do with me personally. They weren't secretly in love with me. Probably. Maybe they had an engineer named Amir.
But honestly, the way IBM went through the dictionary correcting words without always checking what those corrections would break is a fascinating little piece of ViaVoice history.
@brynify With the latest FastSMRW update the user suggestion issue has been resolved. Thanks.
RE: https://dragonscave.space/@amir/117403380740577594
Words like McCarthy, McCartney, McDonald, McKenzie, McIntyre, McLaughlin, and McCormick show the same general ViaVoice/ openevv problem. I'll be updating that document.
Oh wait! ViaVoice apparently can't pronounce McCarthy properly. It's mc carthy in ViaVoice's playbook. ETI Eloquence can handle it properly. Yet another ViaVoice oddity. @nick @fastfinge
@brynify Also if I open FastSMRW and I have no new messages, NVDA will read an older message which might be 6 or 7 messages above the last one. This always happens if no new messages exist upon opening FastSMRW.
@brynify The user mention suggestion feature in FastSMRW is, IMO, broken. For instance, even if I type @bryn, it can't find @brynify @fwoof .space in its suggestion list. Whereas typing @b in Fedra displays @brynify @fwoof .space at the top of the suggestion list, and really quickly.
Folks, no offense intended, and I'm not trying to block anyone. But is there a way in Mastodon, Fedra, or FastSMRW to block or mute posts like this? I'm seeing them more and more lately:
"Delivered X from Y to Z in A: X miles, on time. Earned X achievements, including Just XXX to XXX."
And there isn't even a hashtag we could use to filter them. Honestly I have no idea what these posts are about. They don't make much sense to me, and I don't find them particularly interesting either.
Is there a way to tell my Mastodon client, "I have absolutely nothing against these people, but please make these mysterious achievement reports disappear from my timeline"? Asking for the sake of my increasingly confused timeline.
On the other hand, Fedra does a better job with mention suggestions. As soon as you put an at sign, mention suggestions come up. You should do that manually in FastSMRW, and they appear rather slowly there.
It seems to me that Fedra doesn't provide the same level of sound notifications that FastSMRW does. I'm talking about sounds when, say, someone sends a message or mentions you. So if you're in Fedra and any of that happens, you don't get audio notification.
I'm testing Fedra at work - LOL! But it seems that when I'm using Fedra and others send messages, it doesn't play a sound to notify me. Am I missing something?
Another test.
Test
RE: https://social.lol/@petebrown/117395159325303939
I want to boost this 1 million times!
This stupid Cloudflare "Confirm you are human" checkbox bullshit is quickly becoming as instant a NOPE to your website for me as the "Subscribe to our newsletter" and "Disable your adblocker" popovers. I don't have to prove I'm human, motherfucker. You need to prove you have something worth reading.