Subject:Voice loop problems

Posted by: Jessariah

Date:8/29/2002 7:33:42 PM

Hey all,

I've got some voice loops from an obscure loop/sample disk, and I notice 2 things: 1) pitch shifting takes it right into Alvin & the Chipmunks (understandable) and 2) speeding it up from the original tempo creates a delay-like effect. I haven't had the chance to use vocal loops in the past (except the few ILONA samples that came with AP4), and I know some of you make voacal loops for your projects, so...

Is this just something you deal with when it comes to the human voice, or is there something I can do in Sound Forge to stablize them and make the loops a little more flexible?

Thanks

Subject:RE: Voice loop problems

Reply by: Iacobus

Date:8/29/2002 9:04:16 PM

Almost sounds like these samples weren't ACIDized. Also, as with other types of loops, there's always a boundary between where the stretching sounds natural and unnatural. Time stretching anywhere from 70-115 percent of the original tempo, for instance is usually the safe zone.

I'd say play around with Process>Time Stretch on the menu bar in Sound Forge, but also don't forget about the Special>Edit Tempo and Special>Edit ACID Properties dialogs as well.

HTH,
Iacobus

Subject:RE: Voice loop problems

Reply by: Jessariah

Date:8/29/2002 11:18:56 PM

They're ACIDized - guess it's just one of those things. The Ilona samples I have don't seem to distort as much when they're sped up -- but they also seem to have an effect on them. The loops I have are completely dry...nothing to "hide behind."

Just wanted to know if there was a magic formula out there...

It's funny, depending on circumstances, I've sometimes found that loops sound BETTER shifted than in their original key. Just like certain loops can be sampled at 120 and sound fine at 80, while others sampled at 90 sound terrible at 95...maybe it just depends on the basics of the original wave...

Maybe we need a new book -- "The Science of Loops"...

Subject:RE: Voice loop problems

Reply by: nlamartina

Date:8/29/2002 11:38:24 PM

Jessariah,

Mostly it's just a thing you have to deal with. The human voice is without a doubt one of the most complicated instruments to work with, and your ear will be much more critical about what you hear because hey, you've been listening to this instrument for almost 24 hours a day, 7 days a week for your entire life. You're very accustomed to all the nuances and timbral changes that occur when its pitch moves, so naturally you'll be more sensitive to a computer manipulating it. Really though, it's all about timbre (overtone division, dispersion, and saturation). The voice is one of those instruments that have a timbre that changes dynamically as it moves up and down in it's range, versus a synthesizer, which is static. That's why as a lyric baritone, I sound much different belting a G4 than a mezzo soprano huffing one. It’s the same note, but where as I sound bright, full, and powerful, the soprano sounds dull, low, and dark. Same note, different timbre. At that range, my overtone pattern experiences saturation in the fifth range, from the “i” formant. In the low part of a soprano voice, there’s very little, and it leans on the “o/u” formants. Your brain knows this, and thus you can tell guy from girl, high note from low note, but a computer doesn’t know squat about the human voice and timbre. All it knows is frequency and samples.

Consequently, when a computer takes a voice sample at A4 and shifts it to D5, it doesn't change the timbre along with it, just the frequency along with a lot of interpolation. This is why it sounds “fake”. Sure, in small amounts you won't notice, because timbre changes incrementally too, and isn't missed in tiny bits. It's when you make big leaps that your ear says, "Whoah, what the fark."

Hope this helps,
Nick LaMartina

Subject:RE: Voice loop problems

Reply by: SJH

Date:8/29/2002 11:46:51 PM

Indeed, "the basics of the original wave" ... very important.

What type of source recording, for example. Perhaps on analog tape, than transferred to any number of hard-disk recording systems?

And what did the engineer do that he didn't tell the producer of the loop library?

Was the original file pitch-shifted?

So many variables, and often hard to troubleshoot because you simply don't know the history of your sampled loop.

In your case, your pitch shifting may actually be returning the wave to somewhere near its original state, which would account for the increased accuracy and fidelity. Two wrongs = a right?

I've found that some of the loops I received in my first version of ACID two years ago are sub-standard. Some lack PUNCH, while others are ill-defined, MUDDY. Such is the quality of this relatively new sonic genre.

This week I've been recording my first vocal clips, if you will. Sometimes I can get the loop right and sometimes not. In which case I just make it a one-shot, then draw it in and snap to the timeline. I'm failing as a true ACID-head, but I'm getting my projects finished. And they sound good (with perhaps a little help from my experience as a voice-over actor).

For pitch shifting down my own creations, I speak or sing them a little higher than my normal voice, then adjust in track properties. It seems to sound more natural than if I recorded at my natural pitch. Again, pretty non-scientific, but it works for me.

Cheers, SJH

Subject:RE: Voice loop problems

Reply by: chaircrusher

Date:8/30/2002 1:01:21 AM

I've spent a lot of time arsing around with loops in Acid, and the way that you'll get the best results is to go into the loop's properties, and change 'Force divisions at:' to whole notes. Then manually put in markers at the start of syllables.

If you can get away with it, it's generally better to set the Stretching method to
'pitch shift.' That way there aren't any of the usual timestretch artifacts.

I think that you may simply be trying to stretch your loops too much. There isn't
a program on earth that can play an 80 bpm vocal loop at 120bpm and have it sound
natural. The closest you can come is through phase vocoding, but even if the pitch and timber are exactly perfect, too much stretching is going to sound weird.

If you really really want to use this loop, and you can't use a tempo closer than the original, then another thing to try is to load the vocal as a oneshot or
disk-based, split at word boundaries, and shift individual words around to fit tempo.

Subject:RE: Voice loop problems

Reply by: Jessariah

Date:8/30/2002 8:12:56 AM

Lotta help, info & great ideas here. Thanks to everyone.

Looks like I've got a date with Sound Forge...