Hi, I’m doing a trial of Spectralayers. The main thing I’m looking to do with it is separate male and female vocal parts (singing melody/harmony) that have already been removed from the rest of the mix. From what I understand, “unmix chorus” is meant to be the tool for this, but it doesn’t work at all. Playing around with some of the other tools, “voice denoise” is the only thing which somewhat works, using the “other voice” setting. But it grabs bits of the female voice, then switches to the male voice, and wavers between them quickly at points. It also sounds muffled at some points, and distorted or like a robot at other points. Is this something I’m doing wrong, or is the tech not quite there yet?
Part of what I’ve been doing is separating voices; but not singing vocals…just people talking
The only way I have found in SL11 that is reliable is manual separation…and sometimes you just have to live with the result or see if you can find cover from an identical bit of speech to cover from somewhere else in the recordings…in the event you have it/ have time to find it
I find audio remnants with pretty much all the Unmix modules
like @Phil_Pendlebury often says in his tutorials: SL “is a re-balancing tool” when it comes to unmixing using the current tech
further, I have found when separating voices; sometimes the components can not sound great when soloed, but when you mix it all back together, it can be highly effective
@Unmixing
thanks for the video lesson, it is very kind of you to share your knowledge and process. I certainly learned a few things.
First thing to relate here is that you clearly deeply understand how to read the spectrograph; that is key for anyone attempting to excel at SL. Along those lines, you also understand using SL tools and manipulation of the display…also critical in getting work done in SL.
On the other hand, a great deal of prospects will be looking for the modules to perfectly unmix without having to spend a week of manual selection and cut/paste to another layer. You said it yourself in your tutorial that a lot of manual work is required. The marketing videos kind of imply SL is capable of pretty quick full unmixing. Being efficient at these manual tasks is an art within itself.
Personally, the majority of my work is in mono, so I typically use other colors to guide my unmixing.
I don’t mean to be harsh, yet as far as the quality of the video, you should really at least warn potential viewers about dangerous audio levels. Did you honestly not have time to edit out the speaker/ ear breakers? You are a master of audio separation and you leave in a full blown mic drop?!? Come on, man! I’m very glad I did not listen with headphones. Also, editing video is free with DaVinci Resolve…Videos don’t need to be real time, edit down the waffling…35 min is a pretty long time and you didn’t really show the final result. I mean, things like de-selecting with the rectangle tool…I did not know how to do that and still don’t…I’ll see if I can figure that out today.
Personally, I don’t have enough processing power to leave my FFT size over 3000 smp without freezing…so I (personally) need more power to follow your advice.
Anyway, I think if anything, you presented the case for the OP…these modules are not yet capable of perfectly unmixing without a great deal of manual work.
I will say, that using Magic Wand today for me has been much better than in the past…I’m isolating a dog barking in a scene, and the magic wand combined with brush tool is getting a pretty good result…using larger FFT size though, did not help me…I spend most time at 2500-3072 smp
Can you share the file. Is it mp3 or Wav. 16 or 24 bit. And resolution. The best results we get is on original uncompressed masters where one has no access to the mix but want better mix set to current loudness standard.
Thanks for the replies everyone. I tried to follow along with @Unmixing’s video but got a bit lost after a while. Changing the FFT and resolution settings didn’t seem to help. @Rajiv_Mudgal, I’m using an mp3. It’s a Gillian Welch song that I’m trying to learn David Rawlings’ harmony part to be able to do a cover version of it. How do I share the file? It’s too big to upload and it prevents me from putting a link in here. Thanks!
@lavastudios I hate when people do what you just did. You could’ve saved a lot of time/energy by just simply naming the exact song. It’s kind of hard to help someone when they don’t want to explicitly give details what they need help with.
WHAT IS THE EXACT AUDIO YOU’RE TRYING TO UNMIX? IS IT ON YOUTUBE? CAN YOU SHARE THE LINK?
@Joey_Kapish calm your farm mate, I was asking a general question, thinking I would get a simple answer like “the tech isn’t there yet for separating two vocal parts”, or to hear I was using the wrong tool or skipping a step or something. As I said, this forum won’t allow me to post a link. The full song is on YouTube, it’s called Down Along the Dixie Line by Gillian Welch, from the album The Harrow & the Harvest. I used lalal.ai to isolate the vocals.
@xman_charl Thanks, but I haven’t had any luck at all using unmix multiple voices. It just chops the audio in and out. No separation whatsoever on the few songs I’ve tried it on.
again, I have been doing a lot of voice separation, and my source material is not music…I’m removing crew chatter and wanted speakers on the same mic for re-balancing and I go straight for manual separation; I don’t even try the unmix modules or Voice DeNoise in my work. I find I spend more time un-picking what the modules tried to separate than the time it takes me to manually achieve the results I’m after
I mean, I might try the modules again, but I need a guaranteed work flow
@ctreitzell Wow, I didn’t even realise manual mode existed. I just watched a video on the different tools. I’ve just tried playing around with the frequency tool, but I’m not able to pull anything useful out of it as I can’t tell which harmonics belong to which voice in isolation. I’ll keep messing around with it though.
my test .mp3 file… 2:57 long, 128 kbps
use this for mixing 2 voices, sound forge
use SL to unmix voices
this works well for my needs
have found audio mixes that were junk…
my 2 cents
