Synthesizer V - New voice synthesizer for Windows & Linux (Mac support added)

VST, AU, AAX, etc. plug-in Virtual Instruments discussion
User avatar
pdxindy
KVRAF
16951 posts since 3 Feb, 2005 from in the wilds

Post Mon Aug 27, 2018 10:01 am

quantum7 wrote:Cool technology..,BUT....every time I here “vocaloid” type music it makes me depressed with a feeling that life isn’t worth living anymore. Seriously, I’d almost rather be burned alive than live in a world full of this warped computer-generated distopian vocal-hell. Just my personal opinion though, :)
You already live there... the 'technology' of making a human voice sound 'mechanical' is far more advanced than the other way round.

ENV1
KVRAF
2731 posts since 31 Aug, 2011

Post Mon Aug 27, 2018 12:31 pm

vurt wrote:ill be honest, i never realised how much work went in to just a few lines. almost seems it would be easier to learn to sing :o
gives me a new appreciation for wags efforts.
Yeah, but what i did there is literally doing it the hard way because i used this
talkit.png
to manually create each syllable according to required pitch and speed and then sample it and then put it all together in the DAW for full words/phrases. Basically this was just an experiment to see how far you can get with such an old voice (and it didnt turn out half bad if i may say so myself, especially the parts where i mixed the human voice with the robotic voice), but thats really kinda 'hardcore' compared with whats available today, with something like Vocaloid or any of the other flavors that cropped up lately you could probably do that in 5 or 10% of the time it took me to sample all those individual syllables and the workflow would probably be like 500 times more comfortable too since you dont have to mess around with all those sample files.
You do not have the required permissions to view the files attached to this post.

User avatar
vurt
addled muppet weed
68538 posts since 26 Jan, 2003 from through the looking glass

Post Mon Aug 27, 2018 12:36 pm

ah i see, thanks for the clarification :)

User avatar
SciFiArtMan
KVRian
1186 posts since 24 May, 2006 from Our Amazing Oasis in Space - USA Section

Post Mon Aug 27, 2018 1:53 pm

el-bo (formerly ebow) wrote:That's quite strange to me, as there are moments when her tonality comes very close to that of Susannah Hoffs :shrug:
I agree! In places it brings her image to mind, and the backing vocals sounded good to my ears. But the propensity for her voice to end words with a hard/long E ruins the effect for me. I tend to like having singers end words with less hard Es and Is and more soft vowels.

Like saying the word "time". One way is to say, Tyeeme, and the other is to make that hard "I/E" sound have more of an "ah" sound. Of course if you over-do it, the word could come out sounding more like Tom, than Time. (and no one wants to see Tom in a bottle!). But then that's the beauty of working with a real singer. Or maybe a more adjustable vocaloid robot from hell.

But I also understand this little girl effect is a good thing for the anime audio world, where being a perpetual 12 year old girl is the vision. But I can also see a more mature singer model might have a possibility in a production, even if only a place-holder or demo.

EDIT: After listening to Jaleo's audio clip again, is "she" saying, "It's just a Necromanic Monday"!? This clip has suddenly taken on a far more sinister tone!!! :o
Last edited by SciFiArtMan on Mon Aug 27, 2018 4:32 pm, edited 1 time in total.

User avatar
el-bo (formerly ebow)
KVRAF
16289 posts since 24 May, 2009 from A galaxy, far far away

Post Mon Aug 27, 2018 2:13 pm

SciFiArtMan wrote:and no one wants to see Tom in a bottle!
Even if we did, there never seems to be enough time :wink:

https://www.youtube.com/watch?v=XF8ZgguyQ3Y

User avatar
SciFiArtMan
KVRian
1186 posts since 24 May, 2006 from Our Amazing Oasis in Space - USA Section

Post Mon Aug 27, 2018 6:49 pm

OK, I guess I've been obsessing over that clip of Necromanic Monday by Jaleo. I was immediately struck by what I didn't like about the robovoice, but after giving it all a little more space, I'm impressed with the many things it gets right! Those slurred notes and vibrato kick-ins are really nice! Maybe I need to explore this and other vocal synth apps more and see where the past has brought us. I've been so busy looking to fill a specific need (more ethereal female voices), I've overlooked this robovoice segment of vocal synths.

If more mature voice(s) could happen, and then make it a VSTi, this could be very interesting (maybe this is what Vocaloid is about?)! I doubt any real singers are in danger, but with each step closer these robovoice systems become more and more viable, for "some" use/abuse. I'm also driven to see if some of these shortcomings of the Synth-V voice could be reduced or eliminated with some careful phonetics craft!? Intriguing...

Pelicanomicon
KVRist
50 posts since 29 Sep, 2015

Post Mon Aug 27, 2018 9:11 pm

Just wanted to add that raising the default value of Synth V's "Gender" parameter can help a voice sound less cutesey, same as Vocaloid 1 thru 4. Gender's also good for fine-tuning pronunciation.

Jaleo
KVRist
105 posts since 13 Jan, 2009 from Pontevedra, Spain

Post Tue Aug 28, 2018 12:06 am

SciFiArtMan wrote:OK, I guess I've been obsessing over that clip of Necromanic Monday by Jaleo. I was immediately struck by what I didn't like about the robovoice, but after giving it all a little more space, I'm impressed with the many things it gets right! Those slurred notes and vibrato kick-ins are really nice! Maybe I need to explore this and other vocal synth apps more and see where the past has brought us. I've been so busy looking to fill a specific need (more ethereal female voices), I've overlooked this robovoice segment of vocal synths.

If more mature voice(s) could happen, and then make it a VSTi, this could be very interesting (maybe this is what Vocaloid is about?)! I doubt any real singers are in danger, but with each step closer these robovoice systems become more and more viable, for "some" use/abuse. I'm also driven to see if some of these shortcomings of the Synth-V voice could be reduced or eliminated with some careful phonetics craft!? Intriguing...
Necromanic Monday, hahaha!

I want to share the file so everybody can take a look into it and try the software:

Manic monday.json: https://drive.google.com/open?id=13Wwwf ... UYNemOhhpB

The synth: https://synthesizerv.com/

It's not possible to change the overpronouncing dypthongs but you can mask it with the parameters tension, voicing, and breathness.

I think that overpronounciating is always better than the contrary. I mean, in a future would be possible to add a parameter to morph between vowels for a perfect vowel color, or even make it completely muffled.

Syllabe onset/coda tweaking is needed to improve phoneme timing. Low values to sorthen the consonants and higher values to make them larger.

Another clip looking for a softer voice tone, Nine million bicycles (Katie Melua): https://drive.google.com/open?id=1_F7hE ... hke-8SZeFo

ENV1
KVRAF
2731 posts since 31 Aug, 2011

Post Tue Aug 28, 2018 2:41 am

vurt wrote:ah i see, thanks for the clarification :)
Youre welcome.

It was kinda cool to make such an old TTS processor sing but i dont think id want to do it again because it really is very very laborious.

(That syllable list excerpt i posted is the endresult of what seemed like endless trial and error because for each syllable you have to find just the right balance between pitch and speed and how you write it (e.g. eeo vs you) and also how long the syllable text needs to be (e.g. rayeeeeeee vs ra) relative to the speed set in TalkIt. As you can imagine, this is an extremely slow and tedious process because the only way to check how it works in the song is to sample, transfer samples to DAW, see how well (or not) it works, throw it all away and start over if it didnt work out. (Which i had to do a lot because all you can ultimately do is zero in slowly, bit by bit.) And all in all i have 114 syllables in there (the human phrases, the robotic phrases and some extraneous robotic phrases that arent in the original version) so you can imagine how long it takes to create a complete set of syllables that play well enough together to form acceptable words and phrases. Dont get me wrong though, it was certainly a nice experience doing this and i have no regrets because the result is pretty good for such an old voice. But considering how much work it was i probably wouldnt want to repeat it, if i ever wanted to do something like this again then i would want something that works directly inside the DAW so i can do edits and listen to the changes immediately without having to move sample files back and forth.)

User avatar
vurt
addled muppet weed
68538 posts since 26 Jan, 2003 from through the looking glass

Post Tue Aug 28, 2018 11:31 am

oh definitely, i can understand it would be a lot of work getting each syllable correct, not like youre going to open it and miraculously its the right settings as soon as you type :lol:

some things you have to do once though, if the idea hits you, otherwise its an itch you don't scratch. but i can totally understand its a one off :hihi:

but i am honestly impressed with the dedication involved there, you are a much better man than i in that regard! id have done one syllable and gone "ah eff this, ill do an instrumental" :lol:

so :hail:

ENV1
KVRAF
2731 posts since 31 Aug, 2011

Post Tue Aug 28, 2018 12:49 pm

vurt wrote:oh definitely, i can understand it would be a lot of work getting each syllable correct, not like youre going to open it and miraculously its the right settings as soon as you type :lol:
Exactly, thats what i meant in a nutshell.
vurt wrote:some things you have to do once though, if the idea hits you, otherwise its an itch you don't scratch. but i can totally understand its a one off :hihi:
Yeah, thats pretty much it. Old as it is, there is great potential in the SoftVoice engine itself, but its so-called 'singing' mode is totally useless since you cant control any of the important parameters. (Which seems to go for all apps that use SoftVoice as their basis, and there are many. Willow Pond TTS and ShitTalker for instance. You can always see in the DLLs that there are tons of parameters available but the only ones that are usually exposed on the UI are pitch and speed and some presets like Male, Female, Robot, Martian, etc.) Anyway, since the controls arent there the best you can expect is totally random changes in pitch and vowel-length, all depending on (and changing with) the word and the global pitch/speed settings. So it was clear that wanting to get anything useful out of that built-in singing mode was hopeless. Still i was curious to see how much you can squeeze out of this thing when you just create the syllables in the right pitch and length yourself. And so the project started. And like i said, the result wasnt even that bad. Certainly not stellar, and requiring quite a bit of FX magic to smooth out the rough edges, but for 1984 technology, pretty OK.
vurt wrote:but i am honestly impressed with the dedication involved there, you are a much better man than i in that regard! id have done one syllable and gone "ah eff this, ill do an instrumental" :lol:

so :hail:
Thank you.

Believe me, i was close to saying just that more than once but in the end curiosity and willpower prevailed.

(And whats Radioactivity without the vox. Gotta have them vox.) :)

User avatar
vurt
addled muppet weed
68538 posts since 26 Jan, 2003 from through the looking glass

Post Wed Aug 29, 2018 11:19 am

is it available anywhere to hear?
id certainly be interested :)

although, then it might inspire me to try something similar :o :lol:

and yeah, its actually one of my faves from the autobahn onwards. so cold sounding. very suggestive of a radiated landscape.

ENV1
KVRAF
2731 posts since 31 Aug, 2011

Post Thu Aug 30, 2018 6:00 am

I made it a bit warmer. (From the second pass onward, where the human and robotic voice sing together, it got an arpeggio that adds a nice smooth touch. (Whats the point in making everything exactly the same.) Also the end part is totally un-original but remains 'in the spirit', if you will.)


As to availability, its not published (and it wont be) and im also still tweaking the instrumental parts. But it has a proper intro and ending and is pretty much done so its absolutely listenable as it is. (The levels are still on the low side and no fancy mixing yet but i can run it through BuzMaxi to boost it a bit.) I will attach a render below so you can have a listen. Its not perfect yet but enough to see where im going with this.

ENV1
KVRAF
2731 posts since 31 Aug, 2011

Post Thu Aug 30, 2018 6:39 am

So then, here you go.

I put it in the generic section here on KVR so you dont have to go offsite.

This time the built-in audio player works even for me so it should for you too.

Click blue line below to start playback. (JavaScript needs to be allowed.)

https://www.kvraudio.com/downloads/kvr_ ... 639481.mp3

If the player doesnt work for you just get the file.

Note: There is a second challenge involved here, namely to use only the beta version of Audjoo Helix for all instrumental sounds. So everything you hear except the vox is HelixBeta.

User avatar
wagtunes
KVRAF
19715 posts since 8 Oct, 2014

Post Thu Aug 30, 2018 6:54 am

ENV1 wrote:So then, here you go.

I put it in the generic section here on KVR so you dont have to go offsite.

This time the built-in audio player works even for me so it should for you too.

Click blue line below to start playback. (JavaScript needs to be allowed.)

https://www.kvraudio.com/downloads/kvr_ ... 639481.mp3

If the player doesnt work for you just get the file.

Note: There is a second challenge involved here, namely to use only the beta version of Audjoo Helix for all instrumental sounds. So everything you hear except the vox is HelixBeta.
Not sure what you did but the vocals sound human. Could use some boosting in volume (the vocals) but otherwise, well done.

I'm impressed.

Return to “Instruments”