Those well versed with Australian accents are welcome to check out the latest tuning effort at https://eurpod.com/en_au_tuning.zip - this includes a much more improved pack that solves the 3 phonemes missing from the version initially there. This was part of the SpeechPlayer work in 2016, so whoever tuned these phonemes now gets to have their work live on.
@Tamasg Are you possibly going to do English Ireland, English South Africa, English Scotland and others? Just going off of some of the ones Vocalizer supports.
@mckensie @Tamasg I have a lot of suggestions like this to be honest, especially for dravidian languages, but considering that there are not many contributors, it would be extremely hard. We need possibly two testers for each english flavour or a fundamental hand-tuned phonemes that speechbox can read.
@kaveinthran @Tamasg Speech box might be able to handle ESpeak's scottish ENglish phonemes. Since ESpeak does have English Scotland. I was going to also say if you could get phonemes from Fiona, the vocalizer voice, but that might prove impossible because I doubt you can train vocalizer phonemes but well, good idea for an experiment
@mckensie @kaveinthran @Tamasg These things are literally hand crafted. Vocalizer is separate. Do you know how the phonemes and stuff work Mckenzie? It's all documented. If you want all these languages, take the time to learn how the super complax but worth the playing the phoneme editor is or even handcrafting your own json file by literal hand. That one takes a few hours to get right, oops? When ya have time, ya use it if you're me.
@danestange @mckensie @Tamasg Do you have some resources to learn all this?
@kaveinthran @mckensie @Tamasg Sure, I asked tamas about it over voice. I then later will let codex make its own language. AI is a gift for this stuff too, because it can read the MD file tamas put on the github about the phonemes and stuffly. I'm quite sure there's a phoneme guide there? If not, oops? Codex is pretty good at figuring them out though. My curious question is, can we explain these sliders and thingthe in the app and addon so new users aren't overwelmed? Also, half the sliders in the mobile app are broken.
@danestange @mckensie @Tamasg You connect with him through discord?
@kaveinthran @mckensie @Tamasg Hell not! We connected through stonercloud.net's teamtalk server. I hope tamas feels better today. He's a real great guy and I love to hear him laugh. He's so beautifully great!
@kaveinthran @mckensie @Tamasg Yes, if ya wanna be included, what does ohana mean after all? Pm me with your account and I'll get ya up here. We're chaotic at times, I'm manic adhd. I hope you're all well.
@danestange @kaveinthran @mckensie lol. Or you could feed the 100 KB tuning.MD file to an AI and see how it does with it? ha ha. Everything is always accurately updated at https://github.com/tgeczy/TGSpeechBox/blob/master/Tuning.md
TGSpeechBox/Tuning.md at master · tgeczy/TGSpeechBox

A formant speech synthesis engine with LF glottal modeling, coarticulation, and 26+ languages, written in C++ - tgeczy/TGSpeechBox

GitHub
@kaveinthran @mckensie oof yeah, I don't see Scottish and New Zealand, Irish, South African - all of those have enough linguistic differences that I'd need to know native people in those areas who are willing to sit down and be on like, a voice server where we talk through their speech and tune it together or something. Would be doable but a lot of hard, ear-by-ear type listening and tuning type work. Australian got lucky because the add-on just sat there unused from 2016, never updated, and it was easy enough to compare the main vowels, raise them slightly (like 5%, not a lot) to make sure coarticulation doesn't mess with them, and we got fairly well there now I feel like. And there's way less research in terms of formant tables and targets for various English dialects, so a native speaker + the combination of existing English vowels to tune from ends up becoming an easier path.
@Tamasg @kaveinthran Indian English might get lucky too though, we haven't seen anything from that there though so we'll just have to wait
@Tamasg @mckensie Maybe I am too naive, but in a perfect world, we should already get going an #openSource #linguistic project to document phonemes for every language on earth in human readable and also computer readable language. That extensible library then can be adapted to any TTS we are building. Our worry will be then in tuning the voice and way of emphasizing certain words and phrases.
@kaveinthran @mckensie I think Espeak did this, but their Klatt tables are the standard UK English that SpeechPlayer shipped with. Which is why my only conclusion is that someone (Maybe Mick himself, since I know he was behind the original Speechplayer) sat down with that UK table and listened at least enough to tune out the vowels and get things like the fronting for Kit centralization (the way Australians do their Iy sound) - something I spent hours researching but reverted because for me it would come out more like an Ooh sound all the time.
This looks really interesting though: includes formant targets from 13 accents.
Formant frequencies of vowels in 13 accents of the British Isles
https://www.academia.edu/1357002/Formant_frequencies_of_vowels_in_13_accents_of_the_British_Isles
Formant frequencies of vowels in 13 accents of the British Isles

This study is a formant-based investigation of the vowels of male speakers in 13 accents of the British Isles. It provides F1/F2 graphs (obtained with a semi-automatic method) which could be used as starting points for more thorough analyses. The

@Tamasg @kaveinthran @mckensie for ireish there's this old add on but its mainly ireish Gaelic. but it can do Ireish english the last time i played around with it. https://screenreader.abair.ie/download.html
Screen Reader Voices for Irish by abair.ie

Irish language synthesiser, Sintéiseoir Gaeilge