What's new

Synthesizer V - Vocaloid haters might want to check this

If you want create the background music from mp3 track file, Stagecraft Simple Steems is free for a limited time (give your email to get a free serial). The program decompose any audio into it’s constituent parts. This VST (or standalone) plugin uses the Spleeter algorithm. I have not tried it but it may interest some users.

https://www.stagecraftsoftware.com/products/simplestems/
Has the offer ended? It says $50 and I don't see a free option.
 
I find the new AI retake feature very useful.
Also, maybe I missed something but what is the benefit of the plugin? I’m more efficient exporting/reimporting stuff since the plugin seems constantly put of sync.
 
It was a quick demo to show here the tone and English pronounciation, so there are some out of tune notes and things that could be improved.

I don't use any script. First I create the melody in the DAW, which I import as midi in the Synth V Pro editor, I write the lyrics and from there I start to edit the intonation and adjust the duration and strength of the phonemes.

It's important to determine what's wrong for you in first place, as the app offers the tools to tweak most of those things to your liking:

"T" at end of "last" sounds tacked on: reduce phoneme strengh, reduce breathness.
The noise could be reduced with Breathness parameter.

Him, warning...it never ceases to amaze me how the perception of a syllable changes when a vibrato is out of phase or is wrong.

Fight is short? make it longer! XD

Not sure what you mean with the over emphasized "the"s, perhaps you mean that they start very deep before reaching the tonality?

I've noticed that in other cases at the beginning of phrases with plosives and I think it's more because of Weina's way of singing, but it can be reduced. In general I found Weina's autopitch tunning more pleasant and accurate than Solaria (who is too nervous).



It is still early to draw conclusions but I liked the first contact, the voice is very beautiful and full of nuances. I prefer the Weina's tone and her way of singing, but I'm not the person to leave an opinion about her English. I will publish more demos.

As bonus, another quick and unedited demo just to test the "Powerful" vocal mode + Tension in the high register (forget the flow and the too deep vibratos).

View attachment Ray of light_Weina vocal test.mp3
I agree the singing is gorgeous. If only the pronunciation could be made a little more open. It sounds too muffled for english.
 
The noise could be reduced with Breathness parameter.

Not sure what you mean with the over emphasized "the"s, perhaps you mean that they start very deep before reaching the tonality?
I mean the "the"s are too loud.

Hmm, the noise sounds like a digital artifact rather than what I'd think of as excessive breathiness, but hopefully that will work....
 
I agree the singing is gorgeous. If only the pronunciation could be made a little more open. It sounds too muffled for english.
I've been messing around a bit more and I think the most of the muffled sound is due to the vocal mode. In this demo I used the same track but with different vocal modes applied to each:

Lonestar (Norah Jones) - Vocal mode: Delicate 150%
View attachment Lonestar test.mp3

Lonestar (Norah Jones) - Vocal mode: Firm 50%
View attachment Lonestar (firm) test.mp3

I like the AI retakes system more and more, the timbre and dynamic variations are huge and very expressive. Her tone is awesome.

Anyway, for me Weina English sounds not well defined sometimes, compared to Solaria's perfect, crystal clear pronunciation.

I attached the project file, wich includes a semi-revised Solaria track using soft and airy vocal modes.

Please, comment what do you think of her English.
 

Attachments

  • Lonestar - VI-Control.rar
    37.5 KB · Views: 28
Last edited:
I've been messing around a bit more and I think the most of the muffled sound is due to the vocal mode. In this demo I used the same track but with different vocal modes applied to each:

Lonestar (Norah Jones) - Vocal mode: Delicate 150%
View attachment Lonestar test.mp3

Lonestar (Norah Jones) - Vocal mode: Firm 50%
View attachment Lonestar (firm) test.mp3

I like the AI retakes system more and more, the timbre and dynamic variations are huge and very expressive. Her tone is awesome.

Anyway, for me Weina English sounds not well defined sometimes, compared to Solaria's perfect, crystal clear pronunciation.

I attached the project file, wich includes a semi-revised Solaria track using soft and airy vocal modes.

Please, comment what do you think of her English.
These sound a lot closer to just being an occasional accent, and they're really great in places. The flow is very good, except in "It's dark, and" in the first one.

1:

Issues:

Noisy: "st" of "Lonestar", maybe reduce breathiness or use deesser?

Initial "t" of "tonight" sounds off---partly because it's too soft, but it also sounds like a Chinese accent. Maybe try a different take of the consonant.

The "i" of "fight" sounds like the wrong vowels.

"It's dark and" - "d" of "dark" and vowel of "and" sound wrong, and the flow sounds off (and unnatural).

"I" of "I would give anything" sounds like "of"....

2:

In this one the initial "t" of "tonight" sounds good enough, but the transition into the Final "t" of "tonight" suddenly introduces a very strong Chinese accent---the coarticulation is very different from English, and it's jarring.

"Th" of this and "f" of feeling,

The "i" of "fight" is much better in 2. Sounds great.

Coarticulation of vowel going into the "t" of "It's dark" introduces a strong Chinese accent again. "Dark" has a slight Chinese accent, but the "and" (of "It's dark and") is pretty much perfect in this one.

The "I" of "I would give anything" sounds even more like "of", which is way off. In this style pronouncing "I" like "Ah" is pretty common (Southern / country / Western accent), but "uh" or "of" is just wrong.

"Anything" - that sounds great, love it.

"For you" - the "r" sounds off.

"on me" - might want to tweak the n, though it's not bad.

(It's probably worth mentioning that there are many different accents (and languages) in China, so of course there isn't just one "Chinese accent"....)
 
I've been messing around a bit more and I think the most of the muffled sound is due to the vocal mode. In this demo I used the same track but with different vocal modes applied to each:

Lonestar (Norah Jones) - Vocal mode: Delicate 150%
View attachment Lonestar test.mp3

Lonestar (Norah Jones) - Vocal mode: Firm 50%
View attachment Lonestar (firm) test.mp3

I like the AI retakes system more and more, the timbre and dynamic variations are huge and very expressive. Her tone is awesome.

Anyway, for me Weina English sounds not well defined sometimes, compared to Solaria's perfect, crystal clear pronunciation.

I attached the project file, wich includes a semi-revised Solaria track using soft and airy vocal modes.

Please, comment what do you think of her English.
The delicate setting here is perfect to my ears. As ever i appreciate your examples enormously. This convinced me Weina has superb potential. Thank you.
 
After an endless afternoon of freezing issues with the app, I have something decent enough to show you:

Halo (Beyoncé) - Weina 1.7
View attachment Halo - Weina 1.7 test.mp3
Vocal modes:
Firm, powerful 100%
Resonant 60%
Warm 40%

Halo (Beyoncé) - Solaria 1.7
View attachment Halo - Solaria 1.7 test.mp3
Vocal modes:
Power 50%
Passionate 50%

Edition in both tracks is incomplete as I have a limited time to do this, but the purpose of both demos is to show the timbre richness. The AI pitch/timbre retakes is supercool. The timbre variation is really really huge, not only in the tone and dynamics but it also occasionally adds effects like fry voice or air exhalation at the phrase ends, but also undessired ends. I think I just scratched the surface!.

Weina here nails the tone along the complete register, and again was easier to tune than Solaria.

On the other hand, her English is not as good as Solaria's and the truth is, I haven't stopped to see if some of the faults found could be improved. Some "f"s do not sound as they should, the plosives lack aspiration, long melismas ending in "n", the "n" is too long and the duration must be corrected... to say the most obvious (for me). Some could be corrected in next updates, I guess.

There are no edits in the parameters (tension, breathness, voicing, etc.), except for a few pitch draw corrections in Pitch Deviation when the AI fails completely (always in the same places or circunstances, wathever the voice) and some added falsetto cracks or yodeling effects here and there with no much luck.

Project file attached, enjoy!
 

Attachments

  • Halo VI-Control.rar
    101.2 KB · Views: 37
Last edited:
These sound a lot closer to just being an occasional accent, and they're really great in places. The flow is very good, except in "It's dark, and" in the first one.

1:

Issues:

Noisy: "st" of "Lonestar", maybe reduce breathiness or use deesser?

Initial "t" of "tonight" sounds off---partly because it's too soft, but it also sounds like a Chinese accent. Maybe try a different take of the consonant.

The "i" of "fight" sounds like the wrong vowels.

"It's dark and" - "d" of "dark" and vowel of "and" sound wrong, and the flow sounds off (and unnatural).

"I" of "I would give anything" sounds like "of"....

2:

In this one the initial "t" of "tonight" sounds good enough, but the transition into the Final "t" of "tonight" suddenly introduces a very strong Chinese accent---the coarticulation is very different from English, and it's jarring.

"Th" of this and "f" of feeling,

The "i" of "fight" is much better in 2. Sounds great.

Coarticulation of vowel going into the "t" of "It's dark" introduces a strong Chinese accent again. "Dark" has a slight Chinese accent, but the "and" (of "It's dark and") is pretty much perfect in this one.

The "I" of "I would give anything" sounds even more like "of", which is way off. In this style pronouncing "I" like "Ah" is pretty common (Southern / country / Western accent), but "uh" or "of" is just wrong.

"Anything" - that sounds great, love it.

"For you" - the "r" sounds off.

"on me" - might want to tweak the n, though it's not bad.

(It's probably worth mentioning that there are many different accents (and languages) in China, so of course there isn't just one "Chinese accent"....)
Ooh, thank you for the detailed review, I'll check all that when I have time again.
 
After an endless afternoon of freezing issues with the app, I have something decent enough to show you:

Halo (Beyoncé) - Weina 1.7
View attachment Halo - Weina 1.7 test.mp3
Vocal modes:
Firm, powerful 100%
Resonant 60%
Warm 40%

Halo (Beyoncé) - Solaria 1.7
View attachment Halo - Solaria 1.7 test.mp3
Vocal modes:
Power 50%
Passionate 50%

Edition in both tracks is incomplete as I have a limited time to do this, but the purpose of both demos is to show the timbre richness. The AI pitch/timbre retakes is supercool. The timbre variation is really really huge, not only in the tone and dynamics but it also occasionally adds effects like fry voice or air exhalation at the phrase ends, but also undessired ends. I think I just scratched the surface!.

Weina here nails the tone along the complete register, and again was easier to tune than Solaria.

On the other hand, her English is not as good as Solaria's and the truth is, I haven't stopped to see if some of the faults found could be improved. Some "f"s do not sound as they should, the plosives lack aspiration, long melismas ending in "n", the "n" is too long and the duration must be corrected... to say the most obvious (for me). Some could be corrected in next updates, I guess.

There are no edits in the parameters (tension, breathness, voicing, etc.), except for a few pitch draw corrections in Pitch Deviation when the AI fails completely (always in the same places or circunstances, wathever the voice) and some added falsetto cracks or yodeling effects here and there with no much luck.

Project file attached, enjoy!
These examples are super helpful. Thanks for sharing them with our community!

Scarlet Jerry
 
Is it possible to load SVP files for a voice bank you don't have into Synth V Pro? For example, if I don't get Kevin, could I load a Kevin SVP file to use the pitch data for Solaria?
 
Is it possible to load SVP files for a voice bank you don't have into Synth V Pro? For example, if I don't get Kevin, could I load a Kevin SVP file to use the pitch data for Solaria?
Yes, you can assign any voice you own to each track, but would sing differently and adjustments would be needed. Every time you change the singer, the pitch is calculated again, until you deactivate instant mode. Once deactivated, the current pitch is copied to the Pitch Deviation parameter.
 
After an endless afternoon of freezing issues with the app, I have something decent enough to show you:

Halo (Beyoncé) - Weina 1.7
View attachment Halo - Weina 1.7 test.mp3
Vocal modes:
Firm, powerful 100%
Resonant 60%
Warm 40%

Halo (Beyoncé) - Solaria 1.7
View attachment Halo - Solaria 1.7 test.mp3
Vocal modes:
Power 50%
Passionate 50%

Edition in both tracks is incomplete as I have a limited time to do this, but the purpose of both demos is to show the timbre richness. The AI pitch/timbre retakes is supercool. The timbre variation is really really huge, not only in the tone and dynamics but it also occasionally adds effects like fry voice or air exhalation at the phrase ends, but also undessired ends. I think I just scratched the surface!.

Weina here nails the tone along the complete register, and again was easier to tune than Solaria.

On the other hand, her English is not as good as Solaria's and the truth is, I haven't stopped to see if some of the faults found could be improved. Some "f"s do not sound as they should, the plosives lack aspiration, long melismas ending in "n", the "n" is too long and the duration must be corrected... to say the most obvious (for me). Some could be corrected in next updates, I guess.

There are no edits in the parameters (tension, breathness, voicing, etc.), except for a few pitch draw corrections in Pitch Deviation when the AI fails completely (always in the same places or circunstances, wathever the voice) and some added falsetto cracks or yodeling effects here and there with no much luck.

Project file attached, enjoy!
It sounds awesome. The quieter parts sound a lot better in the first example. The more forceful singing sounds a little better in the second i think. Huge potential. Wonderful sound. Thank you so much.
 
Is Weina significantly worse than Solaria? Or is it a matter of being so new that the best demos haven't risen to the top yet? And I don't mean any disrespect to the people who have posted demos in here. My impression is that it's not user error; the voice just doesn't seem to flow as well.
I was also very disappointed by the sound of the demos I found on Youtube... but here some members are finally showing the potential and qualities of the new VB. Thank you all for the very helpful samples (I was buying all the new voices as a way to support Dreamtonics and because I loved them, but the last releases were not really convincing me, so now before I buy a new one I really want to understand if it's worth...)
 
I don't use any script. First I create the melody in the DAW, which I import as midi in the Synth V Pro editor, I write the lyrics and from there I start to edit the intonation and adjust the duration and strength of the phonemes.
OMG, I've been wishing for a midi import feature in this app... and it's been there all along? I've been painstakenly recreating the melodies all this time...
 
It sounds awesome. The quieter parts sound a lot better in the first example. The more forceful singing sounds a little better in the second i think. Huge potential. Wonderful sound. Thank you so much.
Yes, I agree. I spent more time with Solaria, then duplicated the track changing the voice to Weina, doing some minor editing.

I used Weina Warm vocal mode at 40%, where in the stronger parts Warm must be 0%.

Notice that I was using Solaria's Power vocal mode at 50% only, she can sing even stronger.

While editing I was getting a freeze every 2 minutes or so, which was a pain in the ass.
 
Please stop posting all those great Weina examples, my wallet can't take it! (just kidding, please keep posting them, I really like her voice and I might buy her next month)


Crashing used to be a huge issue for me too with the new update, but it looks like I managed to lower the frequency of them by setting the Power Management option to "Max Performance" and then restarting the rendering system. Might be placebo or just a coincidence, but it does look more stable to me this way...

1658760681475.png
 
New user question. can't find info by search. Can I use synth V like any other vst, use unfinished track on daw and play along with, so I can make changes to music as I go. From what I've seen, you need to Inport track into synth V
 
New user question. can't find info by search. Can I use synth V like any other vst, use unfinished track on daw and play along with, so I can make changes to music as I go. From what I've seen, you need to Inport track into synth V
you can use it as other vst, it will play along with your other tracks in your daw, in the daw that I use, i just need to have synth V vst window open to get it to play along with other tracks. when i finish each section with synth V, i usually record to a new audio track, so easier to playback with other tracks when editing other track in the daw, but i still have the synth v track that i can go back to edit the vocal when i need to
 
Please stop posting all those great Weina examples, my wallet can't take it! (just kidding, please keep posting them, I really like her voice and I might buy her next month)


Crashing used to be a huge issue for me too with the new update, but it looks like I managed to lower the frequency of them by setting the Power Management option to "Max Performance" and then restarting the rendering system. Might be placebo or just a coincidence, but it does look more stable to me this way...

1658760681475.png
Wow, that's solved the freezes for now!

Another fast demo to say thanks:

Never can say goodbye (Gloria Gaynor) - Weina 1.7
View attachment Never can say goodbye - Weina 1.7 test.mp3

In 1.7 the dynamic range in the rendered output is wider(depending the vocal modes used) and I have to set Loudness in Voice Panel/Parameters to negative values to prevent clipping in this case.
 
Last edited:
Top Bottom