NVSpeech Player with phoneme editor version 1.6 now on GitHub as a release.
-
@patricus yeah, that new file at least makes a dedicated phoneme for that sound, so what you can now do is mess with it directly without breaking other languages that way to get the closer sound of it. It should at least be listed inside the language as mapped to another one that's editable so we can tweak it that way.
@Tamasg@mindly.social where to get the new file and how to do it without affecting the damn other languages?
-
@Tamasg@mindly.social where to get the new file and how to do it without affecting the damn other languages?
@patricus oh, it's just that link
Replace your packs folder in the editor with it. https://eurpod.com/synths/packs-PLFix.zip
there are now 13 different phoneme sounds mapped in Polish to play with, and they're Polish-specific, so definitely not going to change other languages.. You have to hit "edit phoneme..." next to the phonemes in language list in the editor. -
NVSpeech Player with phoneme editor version 1.6 now on GitHub as a release. https://github.com/tgeczy/NVSpeechPlayer/releases/tag/v-160
Release notes:
This release removes all direct linking against libespeak and other phonemizer libraries. Phonemization is now performed via external command-line tools (preferably through STDIN), keeping NV Speech Player and the Phoneme Editor fully GPLv2-compliant while still allowing use of GPLv3 phonemizers such as eSpeak NG as separate programs. Functionality is unchanged for users, but the integration is now cleaner, more flexible, and license-safe.@Tamasg@mindly.social I seriously remember messing with that add-on when it twas out... Granted I've never messed with phoneme editing, but that speech is - nostalgic - in my book; Right there with #Eloquence and #DecTalk . -
@patricus oh, it's just that link
Replace your packs folder in the editor with it. https://eurpod.com/synths/packs-PLFix.zip
there are now 13 different phoneme sounds mapped in Polish to play with, and they're Polish-specific, so definitely not going to change other languages.. You have to hit "edit phoneme..." next to the phonemes in language list in the editor.@Tamasg@mindly.social bro I made it, y stil sounds ab bit Chinese but I got it almost working that and the R
I will give it to you
sadly fp is currently having service issues because of code changes. I will just put it into my site
https://posix.live/pl_almost_working.zip
-
@Tamasg@mindly.social bro I made it, y stil sounds ab bit Chinese but I got it almost working that and the R
I will give it to you
sadly fp is currently having service issues because of code changes. I will just put it into my site
https://posix.live/pl_almost_working.zip@patricus sadly I'm stuck in work meetings, might not be able to look at it until the evening, but thank you for the huge help.
-
@patricus sadly I'm stuck in work meetings, might not be able to look at it until the evening, but thank you for the huge help.
@Tamasg@mindly.social some stuff maybe still broken I will fix more later though I think I didn't fix h yet.
-
@patricus sadly I'm stuck in work meetings, might not be able to look at it until the evening, but thank you for the huge help.
@Tamasg@mindly.social ok just a question how to add h?
I'm stuck because the damn editor stops responding whenever I try to make it speak a word with it
I'm new to this stuff but I'm learning hard. I learned what's fricative and stuff like that easily.
-
@Tamasg@mindly.social ok just a question how to add h?
I'm stuck because the damn editor stops responding whenever I try to make it speak a word with it
I'm new to this stuff but I'm learning hard. I learned what's fricative and stuff like that easily.@patricus oh it shouldn't? If it gets stuck or crashes, it could be good to revert to the V5 editor version for now as it might be with how the V6 one introduced the change with Espeak. I'll have to check on that one. Https://eurpod.com/synths/NVSPPhonemeEditor-v5.zip might be less buggy for now, or even V4 if it's really bad. But yeah, something for after work.
-
@patricus sadly I'm stuck in work meetings, might not be able to look at it until the evening, but thank you for the huge help.
-
@patricus sadly I'm stuck in work meetings, might not be able to look at it until the evening, but thank you for the huge help.
@Tamasg@mindly.social ok got h fixed, I used x, modulated it and voila, a bit too harsh but hey, works?
now I will replace ɨ because it sounds Chinese in this engine and use y for some reason is working normally like y, I also noticed I fixed r somehow isn't ideal but I'm learning, ɨ isn't good for y fun experience
-
NVSpeech Player with phoneme editor version 1.6 now on GitHub as a release. https://github.com/tgeczy/NVSpeechPlayer/releases/tag/v-160
Release notes:
This release removes all direct linking against libespeak and other phonemizer libraries. Phonemization is now performed via external command-line tools (preferably through STDIN), keeping NV Speech Player and the Phoneme Editor fully GPLv2-compliant while still allowing use of GPLv3 phonemizers such as eSpeak NG as separate programs. Functionality is unchanged for users, but the integration is now cleaner, more flexible, and license-safe.@Tamasg so i'm trying to add the french a nazelized as in the word "Dans". I don't know which phoneme this goes to, and how to add iti to the language
-
@Tamasg so i'm trying to add the french a nazelized as in the word "Dans". I don't know which phoneme this goes to, and how to add iti to the language
@spacepup For “dans”, the vowel is /ɑ̃/ (nasal A). You usually don’t add a brand-new phoneme right away. Map the nasal vowel to an oral vowel + nasal consonant:
/ɑ̃/ to ɑn (or sometimes ɑŋ, depending on taste)
That already gets you most of the way there and works well in formant synths.
In packs/lang/fr.yaml (or your French pack):
normalization:
replacements:
- from: "ɑ̃"
to: "ɑn"
That way:
• dans (/dɑ̃/) now is dan. True nasal vowels aren’t just “vowel + N”, but acoustically this approximation is very close, and much better than guessing a random vowel. Most engines (and even some older TTS systems) do exactly this. -
@spacepup For “dans”, the vowel is /ɑ̃/ (nasal A). You usually don’t add a brand-new phoneme right away. Map the nasal vowel to an oral vowel + nasal consonant:
/ɑ̃/ to ɑn (or sometimes ɑŋ, depending on taste)
That already gets you most of the way there and works well in formant synths.
In packs/lang/fr.yaml (or your French pack):
normalization:
replacements:
- from: "ɑ̃"
to: "ɑn"
That way:
• dans (/dɑ̃/) now is dan. True nasal vowels aren’t just “vowel + N”, but acoustically this approximation is very close, and much better than guessing a random vowel. Most engines (and even some older TTS systems) do exactly this.@Tamasg but the nazalized a in dans is 1 sound, not that+n, i've been listening to french speakers and that's what i hear
-
@spacepup For “dans”, the vowel is /ɑ̃/ (nasal A). You usually don’t add a brand-new phoneme right away. Map the nasal vowel to an oral vowel + nasal consonant:
/ɑ̃/ to ɑn (or sometimes ɑŋ, depending on taste)
That already gets you most of the way there and works well in formant synths.
In packs/lang/fr.yaml (or your French pack):
normalization:
replacements:
- from: "ɑ̃"
to: "ɑn"
That way:
• dans (/dɑ̃/) now is dan. True nasal vowels aren’t just “vowel + N”, but acoustically this approximation is very close, and much better than guessing a random vowel. Most engines (and even some older TTS systems) do exactly this.@Tamasg but i think i understand
-
@spacepup You’re absolutely right — /ɑ̃/ in dans is one vowel, not “a + n”. Native speakers don’t pronounce an actual n there.
I think the reason people suggest mapping it to ɑn isn’t because that’s phonemically correct, but because of how formant synthesizers work. There is no true nasal vowel tract (no velum lowering, no nasal cavity resonator). A nasal vowel would need different formants + nasal anti-resonance, which we don’t currently model as a separate vowel type. So ɑn is a controlled approximation, not a claim about how French works. Later on we might be able to add more engine params like that though, but it requires deeper level changes to its actual sound. -
@spacepup You’re absolutely right — /ɑ̃/ in dans is one vowel, not “a + n”. Native speakers don’t pronounce an actual n there.
I think the reason people suggest mapping it to ɑn isn’t because that’s phonemically correct, but because of how formant synthesizers work. There is no true nasal vowel tract (no velum lowering, no nasal cavity resonator). A nasal vowel would need different formants + nasal anti-resonance, which we don’t currently model as a separate vowel type. So ɑn is a controlled approximation, not a claim about how French works. Later on we might be able to add more engine params like that though, but it requires deeper level changes to its actual sound.@Tamasg ah okay, i am using your speech player phoneme utility
