The Test: Is Melodyne 5 better for controlling dynamics than a compressor?

VST, AU, etc. plug-in Virtual Effects discussion
KVRist
101 posts since 31 Jan, 2021

Post Sat Jun 19, 2021 12:40 am

I did a preliminary experiment to determine the best way to automate the control of an instrument's macro-dynamics in a song: I took a low C2 pure sine wave at -6db, reduced it by -8db to -14db in Melodyne and TDR Kotleinokov, likely the most transparent compressor I have, then restored the amplitude back to -6db using the gain track control in Ableton, and then compared the results visually for artifacts using a spectrum analyzer. There was very little difference between the original sine wave and the Melodyne one, besides a very slight increase in the noise floor in the most upper frequencies, but at a resolution of near -140db. At -120db resolution, there was no significant difference. The compressor, as to be expected, produced several partials, changing the character of tone slightly.

To make the results more generalizable, I need to repeat the test with more complex, transient-rich material, both monophonic and polyphonic. Despite the overly simplistic nature of this quick test, and assuming that the Note Leveling Macro in Melodyne 5 employs the same process as selectively adjusting one-note, Melodyne appears to be much closer to "traditional" manual automation of amplitude in terms of transparency and control, making it potentially superior to compression in controlling macro-dynamics. Melodyne has the advantage of being able to analyze the full audio clip before applying amplitude changes, whereas compressors react in real-time, dealing with potential distortions from quick attack/release times.

With Melodyne 5's new Note Leveling Macro, one can make both automatic batch and manual changes, raising softer notes, lowering loader ones, or both. Melodyne appears to be better suited for macro-dynamics (note-to-note levels) than compression. Obviously, compression still has a role to play in controlling specific frequencies (e.g., taming subs), changing the character of an instrument through altering its attack-decay-sustain-release (e.g., make a bass more punchy), or creating greater mix density through parallel processing, to name a few of its other uses. While I've gotten good results with controlling bass dynamics using Melodyne, I've yet to test Melodyne's Note Leveling Macro with polyphonic instruments such as a guitar or ukulele. I am curious to see if Melodyne can retain the same degree of transparency regardless of complexity.

The first image is the pure tone, the second one, post-Melodyne, the third post-compression.
You do not have the required permissions to view the files attached to this post.

KVRian
654 posts since 13 Feb, 2013

Post Sat Jun 19, 2021 8:27 am

I don't know the note level feature, but I think it's well known that volume automation is better than a compressor for macro level control, isn't it? The result isn't unexpected to me and I wound not use a compressor to do this kind of levelling in general.

What's the difference between the note level in melodyne and typical vocal rider/levler plugins?

KVRist
52 posts since 19 Jun, 2021

Post Sat Jun 19, 2021 2:49 pm

Paging FabienTDR to verify how apples to apples this comparison is...

Results such as this method creates partials while this other method doesn't makes me think you're doing very different things.

KVRAF
4130 posts since 17 Dec, 2009

Post Sun Jun 20, 2021 12:07 am

compressor = non-linear transfer function.
Non linear transfer functions produce harmonics.
if you would automate a rapid volume change (transition) on that sine wave it would produce harmonics as well.
In order not to produce harmonics while adjusting gain you need look-ahead and fairly slow response.

The only compressor that i can recall from the top of my head that doesn't produce harmonics is FabFilter Pro-C2 in "Mastering" algo.
Honorable mentions; Voxengo Polysquasher3, ToneProjects Unisum.

Yes, for clean nonintrusive dynamic correction clip-gain (what melodyne is doing) is absolutely the better option.

What's worth mentioning is that a good bass player will know how to play into a compressor and unless going for a specific timbre from hard hitting, they will also have a pretty steady level. (so will singers that work with mics).
Don't do macro dynamics just because you think you're supposed to, do it to fix problems or if you're hitting some processing too hard momentarily
Image

KVRist

Topic Starter

101 posts since 31 Jan, 2021

Post Sun Jun 20, 2021 1:47 am

Ploki wrote:
Sun Jun 20, 2021 12:07 am
What's worth mentioning is that a good bass player will know how to play into a compressor and unless going for a specific timbre from hard-hitting, they will also have a pretty steady level. (so will singers that work with mics).
Don't do macro dynamics just because you think you're supposed to, do it to fix problems or if you're hitting some processing too hard momentarily
Good points. I prefer to use as little processing as possible (no compression) while tracking my mainly acoustic music. I like the idea of using Melodyne post-recording to adjust dynamics, if needed, because it doesn't change the ADSR, the original character of each bass note, as does compression, just its overall amplitude. I would use Melodyne's note leveling macro before reaching for something like Wave's MV2, High- and low-level compression plugin. But, obviously, the sound of compression (e.g., MV2) on bass may be desirable in some genres.

KVRist

Topic Starter

101 posts since 31 Jan, 2021

Post Sun Jun 20, 2021 7:37 am

attention ♥ wrote:
Sat Jun 19, 2021 2:49 pm
Paging FabienTDR to verify how apples to apples this comparison is...

Results such as this method creates partials while this other method doesn't makes me think you're doing very different things.
As Ploki mentioned, the creation of partials from compression is to be expected. In another forum I posted at Gearspace awhile back, 'Do all compressors introduce overtones, "saturation"?' (https://gearspace.com/board/mastering-f ... -quot.html), FabienTDR responded with the following:
Whatever modulates a signal in a way or another produces new partials. This is a logical consequence of modulation. These partials can be harmonic, but in real life, they are mostly unharmonic. This is the price to pay for functionality, there's no free lunch.

Our hearing is somewhat more sensitive to distorted steady tones than distorted transients. Note that running a sine through a plugin and into an analyser won't tell much about this story.

Some sources simply don't tolerate added partials or even minutely distorted leading transients. Piano is a typical example, very hard to apply any dynamics processing to.

Try very slow attack and release times in this case, automation, subtle clipping, creative EQing or nothing at all. "Don't" or "Delete and record again" is often the superior alternative when just everything seems to fail.

Return to “Effects”