KVR Audio

emef · Post by **emef** » Fri Feb 13, 2015 11:54 am

some more chipspeech tunage

https://soundcloud.com/michael-forshaw/ ... ma-gave-ya

Valery_Kondakoff · Post by **Valery_Kondakoff** » Fri Feb 20, 2015 12:56 pm

Trying to teach Dee Klatt to speak Russian (around 01:56):

https://soundcloud.com/valerykondakoff/potoki#t=1:56

davidv@plogue · Post by **davidv@plogue** » Fri Feb 20, 2015 3:26 pm

Valery_Kondakoff wrote:Trying to teach Dee Klatt to speak Russian (around 01:56):

https://soundcloud.com/valerykondakoff/potoki#t=1:56

Pretty interesting! In fact we could add some Russian specific logic in there, for Dee Klatt in particular as its 100% synth/formant based.

Valery_Kondakoff · Post by **Valery_Kondakoff** » Fri Feb 20, 2015 7:47 pm

Hello, David!

Glad you answered this message! To be honest, I was planning to post a request for enhanced Russian language support at Plogue Chipspeech forum, but was hesitating a little bit: implementing such a feature was looking very difficult for me (and it still does!)...

The main issues with the current implementation are:

1. The lack a 'soft sign' (palatalization) for consonants at the end of the words. In Sampa/X-Samp it is usually denoted as the symbol ' (apostrophe, ASCII 39), written after the consonant symbol.
2. Then (of course) the 'r' consonant. It is much harder then in English, but when I try to use R instead I'm receiving something noisy, so for the time being I stick with 'r'...
3. I still have not found a way to mark a vowel as accented
4. I'm not sure how to use the $1-4, $6-9 and *1-9 symbols. I was unable to find any reference in Internet after quick search

David, I'm really interested into implementing some Russian language support into Chipspeech. So, if I can assist in any way - please contact me with instructions. I just have sent an email to your support account with a copy of this post (if you prefer email communication). Or I can create a dedicated thread for Russian language in Plogue Chipspeech forum...

Hope to hear from you soon and thank you for your great plugin(s)!

MadBrain · Post by **MadBrain** » Fri Feb 20, 2015 8:43 pm

Hi!

Valery_Kondakoff wrote:Hello, David!

Glad you answered this message! To be honest, I was planning to post a request for enhanced Russian language support at Plogue Chipspeech forum, but was hesitating a little bit: implementing such a feature was looking very difficult for me (and it still does!)...

The main issues with the current implementation are:

1. The lack a 'soft sign' (palatalization) for consonants at the end of the words. In Sampa/X-Samp it is usually denoted as the symbol ' (apostrophe, ASCII 39), written after the consonant symbol.

True, as of yet the closest equivalent we have would be a really fast 'j' or 'i' (something like [j$2]).

2. Then (of course) the 'r' consonant. It is much harder then in English, but when I try to use R instead I'm receiving something noisy, so for the time being I stick with 'r'...

It's true that we could have a better version of [r] - this will require some specific code on our side to properly simulate the tongue roll. For now you could try adding something like [4$1R$14$1R$14$1R$1] at the end of the previous syllable instead (trying to simulate rolling with a series of very fast r-like consonants).

3. I still have not found a way to mark a vowel as accented

This is because chipspeech doesn't support accented vowels (yet). The idea is songwriters typically sing the accents in the text by singing accents in the melody (longer, higher, louder notes).

4. I'm not sure how to use the $1-4, $6-9 and *1-9 symbols. I was unable to find any reference in Internet after quick search

They are nonstandard and are specific to chipspeech. $0..$9 changes phoneme duration in 20% steps (0%...180% of the original length). *0..*9 are used to select non-standard phoneme variants (with 0 being the default variant) that are specific to a voice, but we have very few of those special variants: Dee Klatt has [l*1] (non-velar L, sounds better in non-english languages and can be useful for Russian palatalized L), Bert Gotrax has [e*1] (higher variant of [e]) and [d*1] (faster variant of [t]) but that's the only ones we have.

David, I'm really interested into implementing some Russian language support into Chipspeech. So, if I can assist in any way - please contact me with instructions. I just have sent an email to your support account with a copy of this post (if you prefer email communication). Or I can create a dedicated thread for Russian language in Plogue Chipspeech forum...

Hope to hear from you soon and thank you for your great plugin(s)!

We're glad you like it

dumbledog · Post by **dumbledog** » Sat Feb 21, 2015 12:36 am

Here's why you don't do covers in the same genre I guess

https://soundcloud.com/dumbledog/the-model-demo

Valery_Kondakoff · Post by **Valery_Kondakoff** » Sat Feb 21, 2015 7:53 am

MadBrain wrote:
1. The lack a 'soft sign' (palatalization) for consonants at the end of the words. In Sampa/X-Samp it is usually denoted as the symbol ' (apostrophe, ASCII 39), written after the consonant symbol.
True, as of yet the closest equivalent we have would be a really fast 'j' or 'i' (something like [j$2]).

Yes, fast 'j' or 'i' works well with 'n', 'l', 'm' but not so well with 't', 'b', 'p': [r\at r\atj$2].

2. Then (of course) the 'r' consonant. It is much harder then in English, but when I try to use R instead I'm receiving something noisy, so for the time being I stick with 'r'...
It's true that we could have a better version of [r] - this will require some specific code on our side to properly simulate the tongue roll. For now you could try adding something like [4$1R$14$1R$14$1R$1] at the end of the previous syllable instead (trying to simulate rolling with a series of very fast r-like consonants).

Great example! It seems something like this sounds less noisy: [4$1r\$14$1r\$14$1r\$1] and pretty acceptable. But I still hope for a more tweaking options for r\, R, 4...

3. I still have not found a way to mark a vowel as accented
This is because chipspeech doesn't support accented vowels (yet). The idea is songwriters typically sing the accents in the text by singing accents in the melody (longer, higher, louder notes).
4. I'm not sure how to use the $1-4, $6-9 and *1-9 symbols. I was unable to find any reference in Internet after quick search
They are nonstandard and are specific to chipspeech. $0..$9 changes phoneme duration in 20% steps (0%...180% of the original length). *0..*9 are used to select non-standard phoneme variants (with 0 being the default variant) that are specific to a voice, but we have very few of those special variants: Dee Klatt has [l*1] (non-velar L, sounds better in non-english languages and can be useful for Russian palatalized L), Bert Gotrax has [e*1] (higher variant of [e]) and [d*1] (faster variant of [t]) but that's the only ones we have.

Got it, thank you for detailed explanation and useful examples!

A quick tip for Chipspeech users: don't forget to check the available presets for an examples of usage of different phonemes. Especially useful are: "Games -> Random Swearing", "Misc -> All Vowels", "Misc -> From Vowels" and "Misc -> To Vowels". Personally I was not aware of these and missed them when first trying Chipspeech.

Valery_Kondakoff · Post by **Valery_Kondakoff** » Sat Feb 21, 2015 9:13 am

Valery_Kondakoff wrote:
MadBrain wrote:
1. The lack a 'soft sign' (palatalization) for consonants at the end of the words. In Sampa/X-Samp it is usually denoted as the symbol ' (apostrophe, ASCII 39), written after the consonant symbol.
True, as of yet the closest equivalent we have would be a really fast 'j' or 'i' (something like [j$2]).
Yes, fast 'j' or 'i' works well with 'n', 'l', 'm' but not so well with 't', 'b', 'p': [r\at r\atj$2].

In fact, only 'j' can be used as a 'soft sign', because 'i' is a long vowel.

Valery_Kondakoff · Post by **Valery_Kondakoff** » Mon Feb 23, 2015 10:41 am

There is an edit of Dee Klat singing in Russian (using excellent tips by MadBrain):

https://soundcloud.com/valerykondakoff/potoki#t=1:56

Thanks a lot for the tips, now Dee Klat Russian singing sounds much more authentic. Unfortunately the X-Sampa lines are looking rather gibberish

:

Code: Select all

pO$6 tO ki E l*1e kt4$1r\$14$1r\$14$1r\$1o s$9i ly
ne s$9u$8 tsj$1a v$9O ves$9j$2 O pO4$1r\$14$1r\$14$1r\$1R#
na p4$1r\$14$1r\$14$1r\$1as$9 n$9@ nas# d$9o4$1r\$14$1r\$14$1r\$1 mo$3 z$9i li
na U ke na pe 4$1r\$14$1r\$14$1r\$1e kO4$1r\$14$1r\$14$1r\$1R#

MadBrain · Post by **MadBrain** » Mon Feb 23, 2015 10:38 pm

Nice, it's definitely an improvement yes :3

davidv@plogue · Post by **davidv@plogue** » Thu Feb 26, 2015 3:46 pm

We have just added a new variant of the "Lady Parsec" character voice into our new v1.012 release, dubbed "Lady Parsec HD". Its a non-LPC variant of the voice, and boasts a much wider frequency spectrum and general detail as heard on the original recording sessions we did with the voice actress portraying her.

The new voice is free for registered customers, and you can as usual demo it.
Link: http://www.plogue.com/downloads/#chipspeech

To get some feedback on it we sent the voice to 'Crusher-P', the well known English Vocaloid/UTAU composer/producer, and not only was she kind enough to give us her honest opinions and suggestions, she also made a full song using her! Thank you again Crusher-P!

https://www.youtube.com/watch?v=P4W3f4KDSqw

c_voltage · Post by **c_voltage** » Thu Feb 26, 2015 5:14 pm

Cool

thanks.

Kaboom75 · Post by **Kaboom75** » Thu Feb 26, 2015 7:44 pm

Great demos guys.

I totally understood quantum string theory after using this. Cheers. Would like a vocoder next although I think the one in Live is good enough. I may use vocoding with Chip Speech at some point.

emef · Post by **emef** » Thu Feb 26, 2015 8:47 pm

love the new lady parsec hd voice, she sounds great, spookily realistic

davidv@plogue · Post by **davidv@plogue** » Fri Feb 27, 2015 1:53 am

emef wrote:love the new lady parsec hd voice, she sounds great, spookily realistic

Thanks! This is the first test we did with high quality voice recordings sources.
Since making this even "better" would no longer fit the robot theme, we could spin off the tech for different genres, who knows!

Classical choirs is definitely a thing we are investigating.