Built-in Screen Reader

[Oscar Levin]

I've been playing around with making a built-in screen reader, or "Read Aloud" feature for pretext.  Live demo here: https://math.oscarlevin.com/pretext-testing/ra-demo/read-aloud-demo.html.

I don't want to say too much more, because I'd like your fresh take on how this works without any prompting (trying to find the "desire path" for UI).  But I'd love to hear how it works for you on various OS/browser combinations.

[OHappyDay]

Hi Oscar,

my take on this:

• Looks and sounds good (except "Dr." pronounced as "drive")
• I prefer a little bit lower default speed
• I would prefer to have a choice between a male and female voice. Personally I don't like such a high pitched voice
• Tested on my Android tablet using Firefox

Klaus 

[OHappyDay]

I checked on my laptop:

• Desktop: Ubuntu 24.04
• Firefox:
  • The voice is an absolute no-go. Had to stop the playback immediately. It sounds like a computer generated voice from the seventies
• Google Chrome:
  • Female voice identical to Android tablet
•  Microsoft Edge:
  • Male voice
• Chromium
  • Doesn't work at all. When clicking on the play button nothing happens

Klaus

[OHappyDay]

In Firefox it sounds like this sample:

https://discordier.github.io/sam/

[OHappyDay]

Ok, digged a bit and found the culprit:

• Firefox:
  • It does not come with its own voices (for data protection reason according to Claude) but relies on the capabilities of the OS
  • Ubuntu provides only synthetic voices. I had to install speech-dispatcher-rhvoice and rhvoice-english. After a reboot the voice is now taken from rhvoice (instead of espeak-ng)
• Google Chrome, Microsoft Edge
  • They come with high quality voices built-in so it does not rely on the OS

Klaus

[Sean Fitzpatrick]

I've only got my phone at the moment. I tried Firefox, and Chrome. 

- The Continue button gets cut off on both, even in landscape orientation

- The voice is the same one I sometimes encounter reading text in online videos (I'm not a fan) 

- I didn't realize there was a way to get it to read knowls until I got to the section explaining it. Maybe going in should be there default? 

Overall it's pretty neat!

--
You received this message because you are subscribed to the Google Groups "PreTeXt development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pretext-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pretext-dev/4ed455de-9c1e-47b9-b46f-59e78c398df9n%40googlegroups.com.

[OHappyDay]

An optical hint: (at least on the desktop) the popup for the controls is barely noticeable. Here: white background/black text, the control also comes in white/black. The first time I hit the loudspeaker icon I was waiting for the playback because I didn't notice the control panel at the bottom of the page.

[Mitch Keller]

macOS Safari: 

• The voice is fine. I think it might be the same voice that VoiceOver uses, but I haven’t tested. 
• “Dr. Smith” was read correctly as “Doctor Smith”
• I couldn’t get it to read the second sentence in the Proof knowl in section 1. Example 3.3’s second sentence won’t read either.
• In section 2, “read” was pronounced as “reed”. (I’m guessing this is a common problem with screenreaders.)
• In section 3, when I get to the end of the page, I can’t figure out a way with the keyboard to move back to earlier things to open them and have them read.
• Section 4 appears to be a lie as far as Safari’s implementation is concerned.

macOS Chrome:

• Voice is slightly less robotic than Safari’s. It correctly reads the variable “a”, while Safari’s voice read it as “ah”.
• “e.g.” was followed by a sentence pause breath.
• Same issue with the second sentence of the proof of Theorem 1.1. However, I was able to get the second sentence of Example 3.3 read.
• Do not get the offer to continue to the next section in Chrome either.

One thought, and this maybe should go on -a11y, but what is the best practice regarding the reading order for the image description vs the caption. It feels weird to me that the image description is read and then the caption.

To view this discussion visit https://groups.google.com/d/msgid/pretext-dev/2e1fd3c0-2106-45d7-af3f-5a824586c1aan%40googlegroups.com.

-- 
Mitch Keller
mi...@rellek.net

http://www.rellek.net/

[Rob Beezer]

Firefox on Debian:

Controls become available, nothing happens (nothing read when you press the Play
button).

Chrome on Debian:

* Female voice, pleasant cadence/intonation.
* "Doctor Smith"
* Pause like the end-of-a-sentence after "e.g." (do we need to bug #eg?).
* Can we make a #figure #caption be read at the start? (We place it at the end.)
* Got "read" wrong once ("red" v. "reed")
* No "end of page" announcement.

No more time today, but can help debug later.

Thanks, Oscar - this has a lot of promise!

Rob

> --
> You received this message because you are subscribed to the Google Groups
> "PreTeXt development" group.
> To unsubscribe from this group and stop receiving emails from it, send an email

> to pretext-dev...@googlegroups.com <mailto:pretext-
> dev+uns...@googlegroups.com>.

> To view this discussion visit https://groups.google.com/d/msgid/pretext-

> dev/0a1de395-5d38-46f8-97c9-cdca6aab29b3n%40googlegroups.com <https://
> groups.google.com/d/msgid/pretext-dev/0a1de395-5d38-46f8-97c9-
> cdca6aab29b3n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Daniel Hodgins]

Microsoft edge on Windows 11:

• Works better than the built in screen reader (could read the math and skipped the table)
• It weirdly read one of the proof in a foreign accent and pronounced the numbers in some other language than English
• I played the proof line several times, eventually it went back to English but I am not sure how to recreate this issue

Google Chrome on Windows 11: 

• Works with Chrome built in live captions, and gave a better contrast than the text on the page.

I will ask some of the other grad students at my university to try on iphones/ipads and report back on this.

Thank you,

Daniel Hodgins

[valeriom...@gmail.com]

A bit late, but I tried this feature on my Italian project and it works well. It consistently selects the Italian voice. What I mean is that the voice is chosen based on the project's language, not on the OS or browser language (in my case, Chrome on Windows 11). If I use the same feature with an English project, an English voice is selected.

Is it planned to translate the commands as well? If so, I’d be happy to contribute to the Italian localization.

Valerio