Feature request: Control for the strength of the pitch transitions in Auto mode

I use Synth V, among other things, to create demos for practice purposes for real singers.
With German texts, I forego the search for suitable phonemes and use “la-la-la” instead.
With all the English-language voices that I own, I have the problem that the notes are often “pulled in” from the last note or from a lower tone, i.e. the target pitch is only reached after more than 100 ms, which at 120 bpm is approx. corresponds to a sixteenth note.
In many cases you want this type of intonation exactly, but sometimes it doesn’t fit, and in the case described above, almost never.
image

I was looking for a setting for the Automatic (Sing) mode to reduce these pronounced pitch transitions, like there is for the vibrato influence. Unfortunately, this doesn’t seem to be the case for the pitch transitions.

So far I’ve been making do with manually editing all “disturbing” transitions using the pitch tool, but that requires some effort.

My feature request would be a control that reduces or eliminates the pitch deviations at the beginning of a note without changing the otherwise natural character of the voice.

2 Likes

Did you try - note properties panel/pitch/pitchtransition/offset (and the other associated sliders)?

1 Like

Thank you for the feedback!
Yes, I know these settings, but they are only available in manual mode.
(btw: I use this mode too.)

To preserve the naturalness of the voice, I switched on the global option "Preserve the pitch and vibrato for a note … " for the case “From Sing to manual mode”, so the pitch changes from the auto mode are adopted into the pitch curve and you can edit them.
The values of the sliders you mentioned then overlay the pitch curve.
However, such changes only make sense if you edit the notes individually, otherwise you get unwanted results elsewhere - I tried this for a long time.

If the above option “From Pitch to Manual mode” is switched off, a global influence on all notes is easily possible, but then the naturalness of the voice due to the desired random, minor changes within the notes is lost.

To go back to my example in the opening post:
After switching to manual mode, the pitch curve initially looks like this:

I get pretty close to my desired pitch curve if I delete the negative peaks of the curve below:

However, I get the best results when I edit the pitch transitions in auto mode in the piano roll display with the pitch tool:

As I said at the beginning: It would be wonderful if there was a slider for the auto mode with which you could adjust the delay or strength of the pitch transitions globally for your voice

2 Likes

Have you tried toggling “Improvise Note Attack and Release” in the Voice panel, or adjusting Expressiveness and Enhancement in the AI Retakes panel?

2 Likes

Yes, I experimented with that too.
When “Improvise…” is turned off, this is what it looks like for me:

The AI settings “Expressiveness” and “Enhancement” have a strong influence on the pitch curve, but they do not eliminate the problem. Here both controls are set to minimum (voice: Solaria):

image

If none of those do the trick, generating new AI retakes might be the last option, but at that point rolling the dice is likely a worse solution than the manual correction you’re already doing.

Each voice database will produce pitch curves that attempt to mimic the original voice provider. You can try cycling through the voices you own to find one with a pitch pattern you like, then “lock it in” by changing the notes to Manual mode before switching the vocalist back to SOLARIA.

It’s possible you’ll find a voice database that generates pitch curves that match what you’re looking for, but this sort of delayed pitch transition is something human vocalists do, and that the AI pitch generation has been trained to mimic, so I wouldn’t expect them to disappear entirely.

2 Likes

First of all, thank again for your explanations.

Nevertheless, I wonder whether the pitch progressions really (only) depend on the individual voice characteristics or whether they are also - at least partially - dictated by Synth V.
In my example, if I try out my 5 voices one after the other, but don’t change anything else, very similar pitch transistions (not vibrato!) result:

Kevin:
image
Natalie:
image
Asterian:
image
Saros:
image
Solaris:
image

So I still have hope that it should be possible for Dreamtonics to incorporate a slider that influences the algorithm for pitch transitions. Again: This is also possible and already implemented for vibrato.

1 Like

BTW

Is it intentional or a bug that these AI parameters are not set to the default values when resetting a voice (in the voice panel above)?

Changing a voice doesn’t clear any settings. That would cause a lot of people to accidentally wipe out their work if they just wanted to test out a different voice and forgot to duplicate the track first.

1 Like

I didn’t mean changing a voice - it’s clear that the settings you made shouldn’t be lost.
I meant the “Reset” button in the Voice Panel button, which resets almost all global voice settings (at least those in the Voice Panel), but not the AI sliders Expressiveness and Enhancement, which also have a global effect.

Does adding a small separation between the notes mitigate your issue?

2 Likes

Thanks for the suggestion - I tried it, but didn’t get the desired result for my (admittedly specific) use case.

I would also like to have legato transitions, just not with the note being pulled in, but with exact intonation from the beginning of the note.

That’s why it’s not an issue, but rather a desire for an additional, complementary feature. The current pitch curve is definitely suitable for many purposes.

1 Like

On a first short test, this worked quite well for me. And it is really nice that you can make a batch rectangular selection across “as many notes as you want” and do them all in one go.
I also repeated the same rectangular batch delete on the positive peaks, and I liked doing that in conjunction with deleting negative peaks, because it provides a “balanced compression of both positive and negative pitch deviation”.
(2x view worked better for me than 4x view)

1 Like

Hi there I use Melodyne vocal editing software to quickly solve this problem
I import the rendered Syn V track onto an audio track then use Melodyne to correct any issues globally ie.the whole track , or individual notes …yes the software costs money but it greatly reduces the tedious process you mention. I had Melodyne anyway, but any editing issue can be quickly resolved … pitch transition overshoot , vibrato, volume of notes etc. Just thought I’d mention it.

2 Likes

I’ll try that too, but I didn’t really want to go through the detour of an exported audio file.
But perhaps Melodyne is actually the quickest way to achieve the goal of an unfussy yet natural-sounding vocal line.

I also briefly tried using Antares Autotune as a plugin in the instrument track of Synth V, but that doesn’t work as desired because the pitch deviations at the beginning of a note are in the semitone range (and are not intonated inaccurately),

That’s the point … what helps you best and what is the quickest solution?!

… sounds synthetic, Melodyne doesn’t :wink: … if used correctly. There are superb tutorials on Youtube about vocal fine tuning with Melodyne.

Autotune therefore is just a style tool (wanna sound like Cher?), not a serious editor … at least for my taste.

Hi again I would try it … yes you have to import a Synth V rendered audio file, learn how to use Melodyne but in the long term using this process will become second nature.
Maybe your DAW has the capacity to integrate Melodyne into your DAW… this makes life very easy
( I have Mixcraft 10 Pro )
Anyway good luck :slightly_smiling_face:

just split the note at the position where the pitch transition delays


2 Likes

@Phenocopy Thanks for this suggestion!
Unfortunately, this doesn’t work in a clearly predictable way, and especially only with short notes.
The time required to subsequently split the individual notes is no less than if you edit the pitch curve.

image

Once again: For me it is a feature request, i.e. the wish for a further global adjustment of the voice, as already exists in a similar way for the vibrato strength.
Only Dreamtonics can probably say whether this wish can be realized. Are any Dreamtonics guys reading here?

I’ve already found a number of workarounds, also thanks to the constructive contributions in this thread, thank you again to everyone involved!

1 Like

Dreamtonics does not monitor the forums unless someone reports another user (and even then, it’s mainly volunteer mods, not Dreamtonics staff).

If you want them to see anything, it has to be sent by email to [email protected].

2 Likes