I’ve been extracting and cleaning up voice from a cartoon for some time now and I’m getting pretty good at it after a couple hundreds segments
cleaned up of noise and such. Noise removal effect, equalization, high and low pass filter… I’m getting not that bad with those tools n_n
But there are always some part that are to loaded in music or loaded with annoying sounds for me to clean them up and use them… But I recently
stumbled upon somebody doing the same as me and found some INCREDIBLY well cleaned up segment that were initially clogged up in BG music and seemed unusable.
I have no idea how he managed to achieve it that well, here are the samples :
The first one is the original, the second my attempt and the third his attempt.
What I did to do my sample is first to apply the equalization effect over the whole track at -39Db from 0Hz to 400Hz. Then I copy the track to a new stereo track, remove voice
with vocal remover effect, sample the vocal less copied track with the noise reduction effect and then apply it at : Noise reduc :10, sensibility : 1,65 and 300Hz to the initial track. The result is rather
unsatisfying…
Anybody can give me a hand on this? How do you clean that much mixed up sound? Some other samples of his are even better made…
I doubt I’m missing effects to do it, it would probably be how I use each effects and more than anything else, in WHICH order should I use them.
Here is another one that I can’t figure out…
A small hospital beeping noise clogged up in a snoring sound. The beeping is so high pitch it “should” be removable.
For removing sounds such as that buzz at 1.36 seconds in your first sample, you need a program that can plot and remove frequency traces, (such as Melodyne studio). You can’t do that in Audacity because it is a broadband sound that overlaps other broadband sounds.
“Broadband” means that there is a large (broad) range of frequencies. For example, a flute sound tends to be mostly a few distinct frequencies, whereas the sound of applause, rain, a chainsaw, or that “buzz” has a very much more complex mix of frequencies.
Here is the spectrum plot of a low “A” from a flute:
It would be fairly easy to remove this sound from a mix using a series of notch filters.
On the other hand, this is the spectrum for that “buzz” sound (after your attempt to isolate the voice):
This is much harder to remove.
With programs such as Melodyne, rather than “removing” unwanted sounds, the program can help the user to visualise the elements of the voice (or other sound) that they want to keep. Melodyne is a complex, expensive, but powerful program that I have used briefly, but I can’t justify the cost for what I do. Even with Melodyne achieving good vocal isolation is a long and fiddly job to do well. There is a 30 day demo version of Melodyne Editor here: http://www.celemony.com/cms/index.php?id=demos
I will probably give it a shot on melodyne just for fun, but wow… I looked up a tutorial over youtube with a super simple
song with only guitar and vocals and it already look as tedious as traveling through the sevens levels of hell XD.
I have no idea how he managed to achieve it that well, here are the samples :
It sounds like a particularly good application of Noise Gate.
“I don’t know how they did it” puts you in Urban Legend territory. It’s a completely open question with loads of possibilities including they had access to the original animation files and didn’t separate the sounds at all. It could be ADR and they got their young niece to redo the voice. They could have access to a multi-thousand dollar ProTools install.
Struggling with sound separation is a constant question on the forum and there are no good answers – at least not from us.
There is one other possibility. You got yours from ratty MP3 download and they got theirs from a clean, pure cartoon DVD. MP3 compression damage will kill you any chance it gets. Never do production in MP3.
Yea sorry for the urban legend thingy… I’m kindoff a newb to audio editing you see… so anything past my grasp is legendary to me /)^3^(
I can tell for sure he does not have a contact at the HUB and could never get his hand on the original audio file. Since not all his samples are that well made,
others still have a faint background music and none of them voices are remade as far as I can hear… It should be just a regular recording such as mine.
Pro tool is a possibility… Since everything can be cracked
As for my samples, they come from a looseless .aiff recorded from my computer microphone and the cartoon quality is 1080p.
This should be good enough to ensure respectable quality…
Anyway, tx for your advices I’ll have a look at the noise gate effect.
That’s another pit. People point to a vocal management training video that somebody did on YouTube. Of course it worked perfectly and the performer got stunning results. And you can to, if you use his exact same tracks.
Pick a different track and it all goes into the mud.
There was a product years ago called “Keypex.” It was a very basic volume keyer /noise gate that happened to have very good voice management. It sounded remarkably good. Several radio stations put them on the air for their air talent only to take them off again when it turned out putting the announcers voice over velvety black nothingness is a little disconcerting. Kind of like a robot voice that’s too good.
I clever combinaison of different tools and program can help a bit cleaning sound sample of background noises. But, there is still a type of noise I’m sure I can get rid off and still
can’t find a way to do it. It’s a “aboard a train” noise. I can get rid of the low frequency humming with the equiliser and melodyne for some specific music note or background jingle,
but I can’t find a way to damp the repeating tchuktchuktchuktchuktchuktchuk sound…
I tried with the noise reduction tool, but it only skim this noise when there is no voice being heard and it leaves some pretty obvious trace…
Let your ears have a taste!
Any effect good to isolate looping/repeating noises?
