Pro Audio Files

EQ Ear Training Become a Member

Melodyne 4: Fine-Tune Vocals (Part 2)

Warren: So we’re back here with Stefan from Melodyne. Or Celemony. I suppose it’s the same thing.

Stefan: Yeah, Celemony is the company, and Melodyne is the product we’re most known for.

Warren: We haven’t talked about this, but what else does Celemony do?

Stefan: We have actually a second product, which is independent from Melodyne. You know the different editions of Melodyne, starting with the Essential version, for very low money, up to the Studio version. We also have Capstan, which is…

Warren: And what’s Capstan?

Stefan: Capstan is a tool for removing wow and flutter in a recording, and it’s actually — in the core technology, it uses also the DNA stuff, which made Melodyne become famous for, but it’s a completely different approach on this thing.

And people who use Capstan, they have the problem that they have old tapes, so really old recordings that suffer from wow and flutter, mainly because the tape, over the years, in the archive, it just got corrupted and everything, and so you can restore it. You just put in such a recording that goes, [imitates wow and flutter] all the time.

Warren: Sounds like my singing. [mimics wow and flutter sound].

Stefan: Actually, it’s a funny thing, we started to do this thing that somebody, a Melodyne user had an old recording, and the wow and flutter was going up like this, and he thought, “Well, that’s kind of like some people sing, and you can edit this with Melodyne,” but you can’t, so we developed this Capstan thing, which really solves these problems, and it’s used by people that have these huge archives, and are restoring classical records.

Warren: As you know, there’s very, very few classical pieces where the composer is conducting his own piece, because of course, Beethoven, Mozart, and all of the greats were dead before we had any kind of technology, but there is some early Elgar from the turn of the last century, early 1900’s of him doing like, Nimrod, and some beautiful pieces. You know, Elgar, English composer, heavily influenced by the romantics, and Beethoven’s stuff, and it’s interesting, because I wonder, there’s a lot of tempo changes, which is fantastic, and the pitch remains constant, so you know it’s not a speed up or slow down, but it would be pretty amazing to hear what that would be like if they went through and really cleaned it up, because you know, that — just having the composer conduct his own piece of music is huge.

For any classical music buffs out there, you understand what I mean, because then you really hear what the interpretation is.

Stefan: Exactly.

Warren: Because we’re blessed to read score, but we don’t know Beethoven, when he said to get faster, we don’t know how fast he really wanted to go, and people are obviously a lot more conservative. Especially Beethoven of course, because you know, rubato, and listening to like, Brendel, played Beethoven, it’s so amazing, because he doesn’t care, he just puts in so much expression.

Anyway, I’m off on a massive tangent. As you can tell, I like classical music.

Stefan: We should do a separate thing.

Warren: We should do a separate thing just on classic.

Stefan: It’s really, really great stuff.

Warren: Back to the wonderful world of Melodyne. So what we’re going to do in this one is we’re going to talk about — now, there are some great tools that we all know about, those Pro Tools users will know, things like VocAlign are really great. They go in there, they read all of the transients of the vocal, they take the double, and they line it up.

And it does do a wonderful job, but in Melodyne, you can do a bit of all of that, because obviously, you can align so the vocal is closer. You don’t want it to be so perfect that it phases, but you can get it closer in timing, but of course, you can also work with pitch and all of that stuff. It’s more creative. So why don’t you sort of talk us through what it’s like to work on doubled vocals?

Stefan: Let’s work on this a little bit, and I’ll explain roughly a few functions which you should know as a start to know how to work your way through it. First thing, with this orange button, you get the notes of this lead vocal on the screen, that’s where it goes. You probably would listen to just the lead vocal. So this thing now comes into play, this is the mix slider. You also have it in the plugin, so you don’t want it to work in your DAW and use Melodyne as a plugin, you don’t want to change the DAW mix setup, just to focus on this vocal or this vocal, so you have kind of pre-mix stuff here.

To the left, you only hear the tracks that were put to the orange mode, which all the tracks are showing here. So in this example…


That’s the lead vocal.

Any tuning on it?

Warren: There’s a bit of tuning on the lead, I can hear it.

Stefan: I hear here, and I heard there.

Warren: [singing pitches]

Stefan: Yeah, so it sounds a bit artificial, so that was there before, just to make sure I didn’t do it.

Warren: No, this is a comped vocal, and this is actually what we used in the track.

Stefan: So now, let’s turn on the second vocal, the top vocal, and you can either choose it to become orange as well, so you can edit them. It’s kind of — because they are in unison, they are harder to see the difference between them, and I would start, therefore, to turn them to grey. Oops. Do it like this.

So now they should be overlapping each other, because they are in unison, and I directly see there, there’s a grey note just flat, there’s a grey note too. From first glance, this looks like… Let’s listen to this.


Still hearing only the orange one. Remember, that’s the slider here. I put it back to the middle position, it plays back the orange and grey as well.

[vocals, unison]

And the why, you see that’s just something you need to repair. And by the way, if you turn it further, I’ve learned in the other tracks, the ones that are either orange or grey, which in this example, is our instrumental, I bumped them up a little bit, because when you’re working on a vocal, you should do it in solo, or two vocals in solo. You always need to hear the instruments. Probably, you don’t want to hear them at full level, because you want to focus on the vocals, that’s why we have the slider here, so this would be full level of the playback, and this is blended in just a little bit. So let’s set it up.

[vocals and instrumental]

So that’s maybe a setting I personally would prefer. Now I start working on it. First thing, look for the things that are obviously off. Let’s zoom in here. It’s always important to watch this curve, which is the pitch movement she is doing, and in this note, it’s pretty straight, I’m pretty sure you tuned this. Probably not with Melodyne, because Melodyne wouldn’t make it such a flat line, but this actually had some tuning, so it’s part on the right note, so the lead vocal is correct, and now let’s change the backing vocals.

In the backing vocals, she’s starting pretty low, while the other was straight. It looks to me, it looks like you didn’t turn the double vocal yet. That explains the difference — maybe the lead vocal originally looked like this, also what you didn’t like, that’s why you tuned it pretty hard. So you have one hard tuned vocal which is the lead, and one not tuned vocal which is the double, and this sounds…


If you want to stay with this lead vocal, then you would decide to move the backing vocal up accordingly. I’m doing this here with the Alt key held, so it’ll do it not in certain quantizations thing, it’s doing it freely, and I personally like that it’s taking it from there, so maybe I won’t take it to the maximum. The maximum would be there. If I double click, it’d be at 0 probably percent. Let’s listen to it.


So now it’s kind of spot on. Let’s leave it like this. Another thing I noticed that still I just moved the whole thing up and down, I didn’t move the curve itself. That’s the beauty about Melodyne. If you change the pitch by moving notes up and down, you don’t influence the curve, and remember the curve on the other track — let me just switch again — the curve here is a straight line, you probably edited it like this, and so we would need to make in the other vocal, the line become more straight also. The different strategies of doing it, I would in this case, just really fast, I would probably do it like this with this tool, and I would then build a crossfade here so it still is — there’s still some life to it.


So it’s better now already. Still, I don’t at this particular spot, I don’t like the timing of it. The timing can be improved also. The question is now, would we stay with this part and finish the timing also, or would we disregard the timing for now, and go on through the song, and look for other things where we can concentrate on the pitch? It doesn’t matter actually which running order you do it like this, that’s your personal aside.

So what we would do, I won’t finish the whole song now, but to give you explanation of what I would do, I would go through this vocals and in spots like this where they clash, there I have the — turn on the lead vocal again — so the lead vocal is coming up very high.


That’s what she does. Very nice. And she’s starting, technically speaking, she’s starting much too high. She’s starting over there, a very short amount of time, and then she goes down to the right pitch, and this looks wrong, but to me, it sounds good.

Warren: Oh, it sounds fantastic. It’s one of the things about her vocal that I love, and what I encourage — when I was producing her, I wanted that angst. She was 19.

Stefan: If you’re record vocals, are you aiming for pitch, or are you aiming for expression?

Warren: I’m always aiming for performance. Always expression, performance, because all of the singers I grew up on, the Freddie Mercurys, the David Bowies, amazing singers and incredible ranges. Bowie and Freddie Mercury in particular.

Stefan: So that’s what makes up a great singer is the singer is aiming for expression, and not for control.

Warren: Absolutely. All the time, I use one song, and I will say this again, Ashes to Ashes by David Bowie, where it’s like, [sings song]. I mean, he’s like, three or four different voices within the first 30 seconds. To me, that is a vocal performance.

So she has kind of an angsty, teenage thing in abundance, so I encourage that with her.

Stefan: And you should not do little mistakes. So we shouldn’t encourage the users, if they find something like this, which obviously looks wrong, and it would be easy to correct, I would choose this modulation tool, and I would make it like this. This sounds correct, mathematically, but with this movement, in this very move, I killed this special expression which I wanted. So don’t do it just by looking at it. If it looks strange, it’s a hint that it probably is a little bit strange, but let’s listen back.


Warren: Yeah, “heartbeat.”

Stefan: It’s also, unfortunately, it’s — in the doubled vocal, she doesn’t do the same thing.


Warren: Which would be almost impossible to get that kind of expression.

Stefan: Exactly. So let’s look at it, and she’s coming there, she’s going too far down, which is the first thing I would correct. I would move it up maybe like this, so she lands at the right spot, and she doesn’t start that high. Let’s just, after this one, take a listen.


There’s a timing issue to it as well, we’ll go for this later. So maybe this is okay, or if you want to bring the dubbed vocal more into this expression of the lead vocal, then I would use the tool and go for the other direction like this.

Now, again, compare the two curves. This one goes up, she starts at the F sharp here, and with my latest edit, it started even higher, so I overdid it probably. Just by looking at it probably, this should sound pretty much the same now.


Warren: That’s pretty remarkable.

Stefan: So I brought the same expression that I watched in the lead vocal, which we liked, I brought it in the backing vocals, which you couldn’t do while she was singing it. Again, there’s no strategy of doing, you have to do it like this.

Warren: No, but of all things you’ve shown me, that’s my favorite thing I’ve learned, because that’s pretty remarkable, because I was. I’m trying to get this teenage girl, I mean, she’s still only 21, I did this two years ago, I’m trying to get this teenage girl to be expressive and sound young, because she is young. I’m not trying to remove that from a vocal, I don’t want it to sound like a 40 plus year old guy like me, I’m trying to keep that rawness.

So yes, it’s always — that’s why you can see, I did some correction, but I let so many other things go.

Stefan: What’s important for me at this place that I really want to highlight is for me personally, looking at a vocal is a very important thing. Before I start editing, I really try to understand, “Why is the curve looking like this in this situation? Is it wanted to be like this? Why is it? What does this kind of pitch fluctuation do to the expression?”

And when I say, “I like it,” like in this example, I keep it, and then I look at the other tracks, and I say, “Well, it could be a little bit more like this,” and so I would, in this example, I would edit the same that I just did.

Warren: I think that’s the biggest bonus I can see for me is being able to actually get an incredible performance with expression, and then being able to enhance that is fantastic. Ultimately, I could pretty much take any human being alive and make them sound like they can sing in tune, but I can’t do this. I can’t get expression, and I can’t — that is what makes every vocalist so unique, and yeah, that’s a great, great tool.

Stefan: So at this particular spot, I would then go for the time which we should maybe do right now, but in a real life scenario, I would go through this chorus, and look at the next chorus, which spots I have there, and I pretty much like the performance just by listening to it once. I would assume it would take me maybe ten minutes maximum to do some edits on the lead vocal. There were some spots which I didn’t like for my personal taste, so I would maybe invest another ten minutes.

Warren: Alright, so here’s what we’re going to do. I’m going to shut the [bleep] up, and you’re going to do that.

Stefan: So just give me a few minutes to run through this vocal line. I prefer them to not listen to the double for now, because I want to check the lead vocal, so I just turn the double down.


So I probably find some issues here, this one is a little bit high, just by hearing from it, and I check it, it’s actually only 2 cents, so maybe it was my imagination, but I don’t mind. I put it down a little bit, continue working through the song.


The next thing is the “I” was definitely flat, because she’s starting it very low. If you watch the curve here, she’s starting too low. Much too low. Then take some time. Sometimes, this can be a nice expression, we have a different scenario where we like this again.


This, “I, I would,” if I don’t like it, I would improve it like this. It’s just a matter of personal taste, I just want to show the strategies how you would work on it if you want to work on it. So I would slice the note up there, so remember when I slice note, I showed this earlier, and then it’s recalculating the new note on the screen now, and just by splicing out the note which I did, I don’t change the tuning yet, just the representation of it, but now I see the truth, what’s really happening here. This one’s starting pretty low. Actually, it’s nearly starting on the A, but it’s 40% high, so that’s somewhere in between, so that’s probably what I found disturbing when I stopped the playback here. Just listen once more.


So if I like that she’s taking it like, [sings note], really going up, I would actually even turn it down so it falls into key again, so now it’s really the A. Listen again.


If I like it, that’s good now, and if I don’t like the expression of going like this, in this indie pop thing, I would prefer to leave it like this, if it would be maybe an R&B song, I’d go maybe somewhere like this.


So it’s a matter of taste. For me now, I would leave it down here.


This, “Me,” I don’t like.


I don’t know why I don’t like it. Maybe because it has this very going low to the end. It doesn’t — for me, it doesn’t sound very focused. In her mind, she was already concentrating on the next thing.

So what I would actually do in this thing, I would try to understand why I got a kind of disturbing feeling here. Look for it in the curve, and I see the last part in the curve is what I don’t like, so I move it like this, and what I’m actually doing here, if you grab the normal pitch tool and hover around the end of the note, you change the tempo of the expression, of the transition from one note to the next. So if I do it like this…


The “Me” now becomes much more focused. Actually, while I’m doing it, I still want to give more impression to this “Me,” so I use the amplitude tool, just turn this up a little bit.

And so on. So let’s continue.

Again, I don’t like this being too high when she’s going from this note to this note. Again, it’s a level thing too on this one.


This one, I don’t like, but I guess that’s already been tuned before. Maybe there was something wrong and tuned it already. You can see by the pitch curve here, it’s already pretty stayed. Because it has been tuned before, we did the stem, maybe I can’t do anything on it. That’s a real life situation, actually, you get something for mixing, and somebody had some tuning program on it, and if you don’t like it, you’re kind of stuck. You can, with Melodyne, try to bring back some of the original expression by just kind of over exaggerating what was there. Let’s listen back.


Actually, listen again. It’s interesting. At the beginning, I like this, it’s “let,” and at the end, she’s kind of getting too high, so let’s undo it and compare with the original one.


Somewhere in between would be nice, so what I would do in a scenario like this, I would bring back — rebring back the original expression, which I like at the very beginning, then I would chop the note maybe here, go back to this pitch modulation tool, drag this back to where it was. I liked it.


This “Heart” is the thing we already cared about earlier. It looks wrong, but we decided that this is the expression that is needed at this particular spot, so we leave it like this. I’m not sure about this thing here. This “wake.” Again, it’s a matter of taste. If you like it, so maybe, it sounds for me, it sounds a bit insecure, because again, she’s somewhere in between the notes there. I would actually kind of quantize it, make it the G sharp.


That’s better for me. Maybe even turn it down a little bit more, so the peak of the note is kind of the pitch of this note. That’s what I would do. And so on.

What’s happening here? She’s starting this… while she’s starting low, but again, she’s somewhere between the white key and the black key. I actually like that she’s taking it from down there, but it should be even more down. Something like this…


I prefer this. It’s much more stable this way. It feels for me, it feels more like she really intended to do it like this, but the transition from this first half of this word to the second half is pretty fast, so again, I use this tool and make it slightly slower.


And here, there’s again, this pretty lifeless, maybe it was tuned before too much, and it’s too flat, so with a double-click, I just turn it to one cent offset, and with this modulation tool, I bring in some more of the original life. So something maybe like this.

Okay, download the demo version of Melodyne. It’s for free. 30 days, you can use the full program, Melodyne Studio, same thing as I’m using here. Download it from our website, and use it for 30 days, use it on this material and see what you come up. Maybe this was a big inspiration. What I did in the first course, go ahead, try to do the same thing. Go for the other parts also. It would be really interesting to see what you come up with.

So this was the lead vocals. I would then go to the backing vocals, then we would continue with the other dubbed vocals, if there are any. What I personally do, and I know a lot of producers do it, I know that what Warren gave me in this example was a stem already. He comped the vocals, so there must be outtakes, and probably I would use the outtakes now, probably because there were outtakes because they had more issues probably, but I’m sure that with Melodyne, you could make the outtakes work at least as good, second, third, or whatever doubling vocal.

If the singer already left the studio and we didn’t take more doubles, and we decide later, then we just grab the outtakes, tune them, and then we have our doubles.

Probably, I would also like to do some harmony vocals with it. Again, you should try yourself just with these files. If you can come up with a harmony vocal you like, but when you do it, I want to keep in mind that it’s something artificial we are doing here, we’re creating something that wasn’t there. We are doing already kind of sound design. So again, download our demo version, and see what you come up with.


Warren Huart

Warren Huart

Warren Huart is an English record producer/musician/composer and recording engineer based in Los Angeles, California. Learn more at

Free Workshop Video: Low-End Mixing Secrets

Discover how to make your kick and bass hit hard by cutting (NOT boosting) the right frequencies! Plus, more counterintuitive ways to get fuller yet controlled low-end in your mix. Download this 40-minute workshop by Matthew Weiss, now for FREE!

Powered by ConvertKit