V-vocal degrades quality of vocal clip?

Are you sure you don't have the original clip unmuted? If you do you'll get a slight phasing which may be what you are hearing.

Other than that I'm not sure unless you have one of the settings in V-vocal set to something extreme. I do sometimes get odd artefacts with V-vocal even if not performing any processing but that is very rare and even then it will be one very small part of a clip rather than the whole thing.

Check the original mute status and the settings see if that sorts it.

FastBikerBoy


Are you sure you don't have the original clip unmuted? If you do you'll get a slight phasing which may be what you are hearing.

Other than that I'm not sure unless you have one of the settings in V-vocal set to something extreme. I do sometimes get odd artefacts with V-vocal even if not performing any processing but that is very rare and even then it will be one very small part of a clip rather than the whole thing.

Check the original mute status and the settings see if that sorts it.

Nope, the problem I am describing is definitely not any kind of phasing problem. I made sure that the original track behind the V-vocal track IS muted.


Basically, if I "bounce-to-clip" a V-vocal clip to make the V-vocal clips permanent, and I create a V-vocal clip of that bounced clip, the resulting clip is degraded. Like I described, very hollow and nasally sounding. Very noticeable distortion/degradation.

Multiple bouncing of V-Vocal clips is akin to multiple tape bounces - each bounce will degrade your signal a bit more than the last.

As Kalle asks, why would you want to re-VV an already VV'd clip?

I agree that using V-Vocal on a clip that's already been processed once is bound to cause problems. I guess what you are hearing is re-processed already processed material, a bit like a noise floor being raised but in this case it's artifacts rather than noise.

I have experienced "phasing" sound using V-Vocal.

Bitflipper (member on the forum very knowledgeable) came up with the theory with his experiments that if you donot have headphone leakage, you will get a much cleaner v-vocal clip.

The reasoning is that v-vocal tries to work with event the subtle background noise (the track leaage from headphones) and it causes some phasing and artifacts, etc.

I do not have this problem with Melodyne, howver, which is my tool of choise.

So, if you want to use v-vocal, make sure you have some headphones that seal well (the sennheiser HD280 pros are cheap and very good seal)  and make sure the headphones are not playing very loud so you do not get any headphon leakage.

Thais may clean up your v-vocal clips greatly.

Lance

+1 to Lance's reply - If you decide later you need to do more VV work, start over again with the original clip which is there, muted in the track. Complete ugliness to re-VV a VV clip. And work in small bits. I work with 1 to 4 words/syllables at a time. The parts of a track that have been VVd is a very small percentage by the time I'm through. RB

V-Vocal chops up the waveform into 50ms segments, which is the minimum window needed to determine the fundamental frequency of a waveform across the audio spectrum. 50ms also happens to be the smallest interval in which our ears can detect a pitch variation. Each segment's pitch value is quantized (rounded) to the nearest cent, resulting in a map of pitch changes in a series of 50-millisecond chunks.

You can see this effect by simply bouncing a V-Vocal clip - without any edits - and then recreating another V-Vocal clip from the new audio. Zoom in on the V-Vocal edit window and you will see the individual pitches in 50ms segments.

There is naturally quantization error, which assures that the waveform you bounce will not be identical to the waveform you started with. There are also detection errors in each segment where V-V had difficulty determining the fundamental pitch. You can see this in the zoomed view, where a single 50ms slice may suddenly jump way up, perhaps latching onto a harmonic. Sometimes it looks like V-V is just making a wild guess! These are often too brief to be heard as a discrete pitch error, but they do result in a ragged, unpleasant sound.

It should be noted that all pitch-editing software does this. Re-bouncing a Melodyne clip can do the same thing, and although I've never used AutoTune I would expect it to behave similarly as well. Some degradation is simply impossible to avoid.

So why does Melodyne seem so much more transparent than V-Vocal? It's because of better pitch detection. Melodyne is better at throwing out the statistical anomalies that V-Vocal just accepts as legit micro pitch changes but are really detection errors. Melodyne will interpolate disparate pitch segments and can even make a reasonable guess in breathy parts where V-Vocal just gives up. But even Melodyne has limits, and a badly-recorded vocal can confuse it, too.

Another issue with V-Vocal is that its threshold for detection is higher than Melodyne's. I've forgotten now what my experiments suggested that threshold to be, but it was fairly high, like -24db or so. Consequently, quiet passages appear to be skipped over (no white horizontal guide lines). Melodyne is happy to detect much lower levels, so it's less likely to give up mid-phrase - but it still gets confused in quiet parts and long tails. Starting with reasonably hot vocal levels (> -20db) is a good idea regardless of what pitch-editing software you use.

The bottom line is that pitch-correction causes signal degradation. Count on it, and form your editing strategies accordingly.

Some thoughts on pitch-editing best-practices:

1. Start with clean, loud, unaffected mono vocal parts. Avoid distortion, clipping, headphone bleed and other extraneous noise that will confuse the editor's pitch detection. Clean up sibilance, trim leading room noise and mute silences before editing pitch. Never record with reverb, delay, chorus or any other effect, and always record in mono.

It can sound really cool to record vocals via a mic in front of a PA speaker, but if you do that, resign yourself to not doing pitch correction later. Singers enjoy hearing reverb and delay while they're singing, so accommodate them in the headphones - just don't record it that way.

2. Do not apply correction to any parts that don't truly need it. Don't ask V-Vocal which parts need it, it will reply with exaggerations and lies. Close your eyes and listen.

A general principle of recording is that anything you don't touch is something you can't f*ck up. The kinds of records that get engineering Grammys have usually been made with a very light hand.

3. Isolate the phrases that clearly need editing by splitting clips, so that you're only editing small sections of the performance. Same principle as above: you can't make it worse if you just leave it the f*ck alone.

4. Set the Formant knob to zero. V-Vocal totally hoses formant-tracking, one area where Melodyne wins hands-down. Setting the formant adjustment to zero is just telling V-Vocal not to even try. There are exceptions to this rule, but only if you're using V-V as a special effect such as chipmunks or deep, sinister voices.

5. Always adjust by individual words and phrases, not entire clips. Highlight the specific section you want to edit, leaving out leading and trailing pitch-less consonants from the data selection. Short consonants such as "s", "k" and "t" have no pitch, so leave them alone when possible.

6. Always make adjustments by ear, not by eye. Pitches don't have to be smack on to the cent to sound good! If in doubt, close your eyes and listen. If it doesn't clearly sound out of tune, leave it alone.

7. Restrict your pitch changes to one or two semitones. Beyond that, you're really asking too much from dumb software. And really, if your vocalist is that far off he or she simply hasn't learned the part yet and needs to rehearse more.

It's tempting to use V-Vocal to fabricate vocal harmonies, and I admit to having done so myself for parts too high for me to hit, but vocal parts generated this way rarely stand up on their own and can only be useful when buried in the mix.

8. Never put your pitch correction software on auto-pilot and let it make all the decisions for you. Always make every correction by hand. This is really just a corollary to the above principles, but vendors like to advertise the ease-of-use angle and post YouTube videos showing how to fix pitch with two mouse clicks. Don't fall for it!


Some day, I'll have to write these down and come up with two more bullet points so I can call it a proper Top Ten List. Feel free to help me out in that regard.