Karoryfer releases Hster, a vocal sample library in Proto-Indo-European

[DSmolken]

https://shop.karoryfer.com/products/hster

For some reason (feel free to speculate wildly what it might have been) I recorded an opera singer singing in a reconstruction of a language which was spoken some 5000 years ago, and made a sample library of this. It works much like our sample libraries in Ewe, but Proto-Indo-European is the opposite of Ewe in this way: it has lots of consonants both at the starts and ends of syllables, and it has only two vowels.

Walkthrough:



And the singer's favorite out of the first demos - more demos on the page.



$19 intro price until August 15th, $39 regular price.

[Michael L]

For some reason (feel free to speculate wildly what it might have been)

Well obviously this language is like the waveforms of a synth.

It's the set of basic sounds that humans used to compose all ~100 modern indo-european languages.

I will further speculate that you could design a cyborg page that would let us create every other language sound from just this one sample library!!

If anyone can do it, you can

[DSmolken]

Good theory but no, you'd need a lot more data to be able to derive all the other languages - this one doesn't even have an a vowel.

[Michael L]

OK, so my "wild speculation" is sadly NOT why you "recorded an opera singer singing in a reconstruction of a language which was spoken some 5000 years ago" but it IS a good reason! And with only Estonian (45 vowels) & Lithuanian (47 consonants) opera singers you can get close

[DSmolken]

The opera singer part is really straight up practicality: she had time, and singers who are trained for consistent pitch and dynamics make the editing process faster.

[Michael L]

I am sure the feeling was mutual. You are probably the first man she met in a long time who truly appreciated her talent of singing Proto Indo European.

[DSmolken]

New demo! This one's even more sci-fi... not necessarily what I would have expected for such a non-retro vocal, but hey...

[DSmolken]

Last 24ish hours of intro price... I'll set it to full after breakfast tomorrow.

Kinda shocked how many people actually need to score music for prehistoric settings, and this really does solve a problem they had. Though, of course, a lot of people don't really need-need it but bought it because they think it's cool, and there's nothing wrong with that, either.

[GordonnS]

I'm very, very, late to this thread, but recently bought Hster, so this seems a place to ask.

Is there a best method for sending the CC55/56/57 messages for the sylables and ending, please?

[DSmolken]

I hope somebody else pitches in... tbh I haven't used anything except the vowels from this vocal since releasing it. I produce music for singers, so I use samples of my own instruments all the time but vocals not much, and very rarely syllables. But from what I remember, for the walkthrough and demos I just picked one syllable (they should all have the same onset timing), played the part in from a keyboard live, and then added the syllable-swapping keyswitches or CC messages after the fact.

But while we're here, and speaking of producing music for singers, I produced something for this very singer just a couple of months ago, using also all my own instrument samples. And now we might even put together a proper lineup and start playing live.

[GordonnS]

But from what I remember, for the walkthrough and demos I just picked one syllable (they should all have the same onset timing), played the part in from a keyboard live, and then added the syllable-swapping keyswitches or CC messages after the fact.

That's presently my best guess as a procedure, too, though reversing it might also be a candidate if trying to create 'proper' speech. I have wondered about using a matrix controller, though that may need MIDI-message translation and be awkward to use in real time.

[DSmolken]

Realtime is... yeah, probably not really feasible. Some choral libraries and Chipspeech are playable in real time if they have the syllables pre-programmed in advance and automatically advance one syllable during a break between notes... but that takes pre-programmed syllables. To do it really real realtime still seems to be the realm of living human singers, I think.

[GordonnS]

Yes, I've also now concluded the same. PIE is too far removed from my experience and has too many complex word structures to do live. I will definitely need to edit it in a DAW oe sequencer. I've been trying to work out if some form of scripting might be possible, but I'm having enough trouble just understanding the language, not helped by slightly varying text/phonetics methods.

I've long been curious about how our languages evolved, so I'll definitely be studyting further. It's heavy going for me at present.

[DSmolken]

Ha, it can be quite the rabbit hole. It's funny how languages like Ewe or Japanese are relatively easy to synthesize because there are practically no consonant clusters, but a pain to sample because they have a lot of vowels so if you're going to record true legato for the vowels, that's a whole lot of recording and editing. PIE is the opposite - lots of consonants, often in big clusters, and four vowels as spoken, but that's two short and two long versions of the same two vowels, so legato is quicker.

All this started out with somebody in another forum requesting a sampled Sanskrit choir, me thinking "why not go all the way back to Proto-Indo-European" very much as a joke, and then reading up on it and realizing it's very much doable. In a way, I trolled myself into doing this.