I’m trying to isolate live vocals from a video camera that was set up near the singer.
How would you go about it?
The things that have to be removed in my case are the crowd, the playback track, including the background vocals and noise, clicks et cetera.
Also, do I remove hum, noise, clicks and reverb etc at the very end of the processing or in the beginning?
I would love to know how you would tackle the session?
I’ve just unmixed the song for the vocals so far and used the identifying multiple voices to remove the background vocals but that doesn’t seem to work well.
yep, I’ve never gotten useable results with Unmix Chorus or Unmix Multiple Voices in SLP
And separating multiple people speaking at one time in one mic is very challenging manually. Singing is a different story.
Really what you need is a clean vocal to use as a selection source layer, for example with interview/ documentary recordings where the contributor wears a lav and you want to clean up the boom. I reckon you don’t have that, tho.
IME, Buzz/Hum/Electrical noise is typically the easiest component to remove/reduce.
I shouldn’t think you could get solo-worthy stem isolation without a great deal of artefacts, but you should be able to rebalance. The years I’ve been using SLP, I don’t think I’ve ever used Unmix Crowd Noise.
and thus a lot of manual work ahead of you, I’m afraid :-/
depending upon what you hope to yield for your efforts
and likely best you can really hope for is a re-balancing
clearly a lot of complete NR like camera capturing telephonica (i.e. HF hyphenated lines) and other buzzes and hums (appears as kind of a straight, horizontal line)…multi-band and broadband unwanteds are much more difficult
just running modules is likely not going to yield the results you are after.
Do you have a reasonable grasp on selecting human voice in the spectrogram? I find magic wand with good settings can be much more effective tool than the freq or harmonic selection tool…that said, knowing when to select hi freq consonants verses lower freq vowels and having a grasp on broadband transients are the basic abilities needed.
If it was me I’d look into other automatic stem separation tools before diving into the SLP time drain of manual selection. Get the best result you can find that way and then bring back into SLP for fine tuning
It sounds to me like you have a live recording from a physical mic attached to a camera? A literal “Camera Mic” situated at a distance from the singer/band… maybe in the audience even?.
If so, in my experience you might not be able to pull off what you want. Tools like SL work best when the source audio is basically good already, and we are just tweaking the last few percent to get it perfect.
A recording from a camera mic is rarely useful other than syncing with another better recording. The signal to noise ratio ( ‘noise’ being the band, the audience, ambient sound, camera noise etc ) is usually way too low to be useful and still sound ‘good’ after treatment. It’s just physics.
In your scenario I think a clean feed from the mixing desk would help … or even better if it’s single you want to focus on an isolated recording from the vocal mic. This could from the mixing desk or from a stage mic split. If you had this you could probably clean it up nicely in SL… then remix it back into the main recording.
That’s a more complicated recording set up, but it is ‘how it is done’ in the industry.
I’ll still be using SpectraLayers for the basic noise removal but lalal just slayed it.
I didn’t have to tweak anything and the echo removal is also top notch. Whatever lalal gave me is way better than what I was able to achieve with Spectralayers.
SL should drop the line on steam seperation, because there are way better models, programs and tools out in the market.
SL should be (imho), what it wants to be: The, by far, very best spectral EDITOR - not more, not less.
SL becomes rediculous, if stem seperation remains its primary goal. Not even mentioning the trouble with Apple hardware…
Just saying!