Collect Element only collects Label or custom text, ignores Pronunciation?

[maacsinreality]

Hi
Since it looks like Google have zero interest in ever fixing the Chrome switching voices bug for PWAs and, despite your best efforts how ever many months ago, Firefox page drawing remains too slow on low-end Android hardware - thanks RAMageddon shortages and price hikes - together with its continuing weird 'overscan/underscan' full screen filling troubles for it to be an option for a daily driver, I've resorted to getting the Spanish (Es) Google TTS synthetic voice to phonetically pronounce the limited selection of words in the [ENGLISH] category as naturally as possible (to my native English-speaker ears). The same workaround as I'd employed for years with LetMeTalk to try to make that slightly bilingual.
Obviously far from ideal, but it works of a sort.

Speak custom text has been the only way I've found to retain correct English spelling as the Label - despite the user not being able to read anything - while getting the Collect Element to output the tweaked phonetic attempt of the Spanish voice.
For example, Label [ I ] has to become Custom text [ AI ] otherwise its output sounds like 'EE'.

Isn't that what the seemingly slightly more flexible Pronunciation field is designed to achieve, but only seems to work with Speak label? Can the Collect Element be made to use Pronunciation, when different from Label?

Thanks/Gracias

[klues]

Yes, it's really a bummer that Chrome isn't fixing the voice switching bug.

The pronunciation field in the general element properties was introduced as a kind of shortcut for the cumbersome setting up of speak custom text actions for fixing pronunciations.

So currently it's like this according to my tests:

• speak custom text action:
  • when adding to bar, speaks the custom text
  • shows the custom text in the bar, not the label
  • speaks the custom text in the whole sentence in the bar
• with pronunciation field set:
  • when adding to bar, speaks the custom pronunciation
  • shows the original label in the bar, not the pronunciation
  • speaks the original label in a whole sentence

So if I understand you correctly, the only thing you would like to change is that for the pronunciation field the last point speaks the original label in a whole sentence is changed to speaks the pronunciation in the whole sentence while keeping the original label shown in the bar. Correct?

[maacsinreality]

Correct.
Although I hadn't extended the thought process to whether Label or Pronunciation should display in the bar since current usage is solely pictograms/images in it. The cells themselves have Label so yes, I'd say Label should appear when that option is selected exactly as your understanding, not the odd combination of characters usually needed in Pronunciation.

Thanks

[klues]

Great, implemented it. Please test here: https://gridbeta-github.asterics-foundation.org/klues/851-pronunciation-in-collect-element/pr

[klues]

by the way - I think I've just found a workaround for the Chromium bug - see #856 🥳

[klues]

since I'm quite sure it works as intended, I've already released it with https://github.com/asterics/Asterics-AAC/releases/tag/release-2026-09-18-08.56%2F%2B0200

But if not, please comment again.

[maacsinreality]

While it does appear to work as originally conceived, the unexpected but very welcome introduction of #856 means that there are more combinations to consider. Yes, unfortunately, I would like, if possible, to maintain the Google TTS Spanish (es) voice's attempts at reproducing English words as well as having the English (en-gb) one.
I shouldn't be negative, but knowing Google's ability to do whatever takes their fancy in the name of so-called "Security" enhancements, then I'd like retain the fallback position should your new-found "workaround" suddenly be stopped from being one with a future Chrome update.

Principal using Spanish (es)
Label:⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀ LEAVE ME ALONE
Pronunciation:⠀⠀⠀⠀⠀⠀LIV MI ALOWN

Translated using English (en-gb)
Label (en-gb):⠀⠀⠀⠀⠀⠀ LEAVE ME ALONE
Pronunciation (en-gb):⠀______________

The Collect Element speech appears to focus solely on prioritising Pronunciation, wherever it finds it, which means the English (en-gb) voice incorrectly attempts to say "LIV MI ALOWN" not its Label "LEAVE ME ALONE".
No time to run through permutations, like if both have alternative Pronunciations (for whatever strange reason) which one might be chosen, etc.

Thanks

[klues]

Ok, I think I know what caused this problem - please test a new fix at: https://gridbeta-github.asterics-foundation.org/klues/851-pronunciation-in-collect-element/pr

[maacsinreality]

The workaround was to ensure that the Pronunciation entry was never left blank - implementation of which would've extremely laborious since Element labels columns unfortunately can't be cut'n'pasted into their associated Element pronunciation ones in Translate grids. Not that that is something grid designers might be wanting to do very often, is that extra flexibility a 'quick fix' possibility?

Anyway regarding this actual fix, then it appears to work as hoped - probably negating the need to ever copy Label to Pronunciation en masse, nor vice versa - but during my 'button-bashing testing' of it another Always wait Separately (and perhaps more importantly for other users since it also affects Speak label as originally implied in #827) 'instability' exposed itself that is better documented by Reopening #846, when I get a chance to try to summarise the repeatability factors (not necessarily involving this fix at all).

[klues]

ok, but the pronunciation thing now works as intended, right? So I can mark this as ready-to-release?

[maacsinreality]

Seems to, so sign it off.