The idea does sound good, I want to add though that we are kind of depending on having a open source library the provides audio rendering of IPA markup. Is anyone aware of one? KSiebert (WMF) (talk) 13:58, 19 January 2022 (UTC)[reply]
I support this solution. When I read biographies in other Wikipedia, I want to hear how the name and surname are pronounced. IPA is unreadable to me. A reading machine would help a lot.--Rosewood (talk) 09:46, 21 January 2022 (UTC)[reply]
Hi Eptalon. Yes, mostly with tonal indication, but it is not such a big deal, there is several systems but no overlapping between them. Something else: A language with a large diffusion could have more than one way to capture the phonological representation, I mean the most systematic one despite local changes. This could still be conflictual between contributors when a phonetic change is typical in a large area such as a change on the sound of the letter d in Canadian French. Well, that was a serious answer. I hope you got the point (I am kind of a linguist). A shorter answer could have been: Yes, one IPA is a beer, the other one a script Noé (talk) 13:33, 25 January 2022 (UTC)[reply]
It should be able to, to some extent. It would help if there is a consistent markup for IPA (don't know if this is the case). As discussed above, IPA can be written a bit differently and with varying levels of detail. Having a TTS that can pronounce all possible IPA may not be realistic and some conversion will likely have to be implemented. E.g. when I did a quick test now I had to convert to /ˈbɝmɪŋəm/ for the TTS used by Wikispeech. Sebastian Berlin (WMSE) (talk) 09:07, 31 January 2022 (UTC)[reply]
Can IPA actually do this? The enwiki IPA help pages are full of indications that the same IPA is pronounced differently in different accents or dialects of English, warnings that the symbols aren't directly equivalent in other languages etc. If it can work reliably, this would be a useful feature, as IPA is incomprehensible to most readers. But we don't want a system that pronounces every word in (say) a mid-Atlantic accent, as if that's the only correct pronunciation. Modest Genius (talk) 20:49, 31 January 2022 (UTC)[reply]
Yes. See the linked proofs-of-concept, on the Phabricator ticket. Also note that at least one of them offers a choice of voices. Even so, better to have "every word in (say) a mid-Atlantic accent" than no audio pronunciation at all. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits21:20, 4 February 2022 (UTC)[reply]
Neutral Forgive me if I'm being arrogant here, but I don't feel IPA is too hard. I worry an automated voice could be difficult for some, especially with dialect differences. Celerias (talk) 11:15, 29 January 2022 (UTC)[reply]
Support Is not big problem to read IPA (because characters are letter-like), but is problem to write it and understand every nuance. JAn Dudík (talk) 13:03, 31 January 2022 (UTC)[reply]
Support Would be fabulous, I imagine it would be a welcome feature for the wider group of readers as well. ~ Amory(u • t • c)20:51, 2 February 2022 (UTC)[reply]
Support I would also prefer human recordings instead of audio rendering, but issues of scale aside, I think doing something like this (1) would provide an immediate benefit to many, many users on many, many articles and (2) might stimulate people to say to themselves, hey, I can record a better pronunciation than that -- and then hopefully they'll do that and upload it to the Commons. Gaurav (talk) 03:22, 11 February 2022 (UTC)[reply]
Noé (talk) 13:33, 25 January 2022 (UTC)
Support --Nachtbold (talk) 18:35, 28 January 2022 (UTC)
Neutral Forgive me if I'm being arrogant here, but I don't feel IPA is too hard. I worry an automated voice could be difficult for some, especially with dialect differences. Celerias (talk) 11:15, 29 January 2022 (UTC)
Oppose --Ciao • Bestoernesto • ✉ 03:31, 6 February 2022 (UTC)
