Please simplify song unmix & improve workflow!

SpectraLayers
1

Hi guys,

The song unmix feature is amazing, but it could really be better in terms of workflow.

Here are my suggestions:

Combine the separate process into one:

So, just give us check boxes for all possible layers:

bass

drums

snare

kick

brass

backing vox

etc

etc

So, it shouldn’t be in a two part process where you first unmix the song, then unmix the drums.

Secondly, please output the resulting stems to the same folder as source file, naming the folder :Let it Be - unmixed" or whatever.

Each unmix file should be prepended with the WAV title, so: “Let it Be - bass.wav”
But we also need an option to input the text that will be prepended (the reason for this is explained later in this post).
So, automatic prepend would be: “Let it Be - v2 FINAL - 12.12.24 - with comp no limiting - bass.wav”
But we would have a text box to change that to: “Let it Be - bass.wav”

Also, it should generate a text file in that folder listing the stems.

AND, and a later date, when the artist sends a new version of the mix, you can drag that text file to the batch process window, and that will automatically recall all the settings for the original unmix (and whatever other batch processes).

It would be great if SL could also detect if the new mix has the same or a different length from the previous version, throwing up a warning that you need to edit the new version before running the process.

And it should also rename all the previous versions, with “OLD”, so then it becomes: “OLD - Let it Be - bass.wav”

So then your source folder will now contain two folders:

Let it Be - unmixed

OLD - Let it Be - unmixed

You might think that should also add dates to the file or folder names, but it’s VERY important that this doesn’t happen. The new unmixed stems MUST have the same name as the original ones, for this reason:

Imagine that you’ve worked on (in my case, mastered) a song, perhaps done quite a bit of editing or object based FX. Then the artist wants to send you a new mix.

If the new (SpectraLayers) stems have the exact same titles and folder structure, then when you open the project in your DAW, the DAW doesn’t know or care that the stems are different (new) files, as long as they have the same name and length. This would save A LOT of time in revisions. Otherwise you need to manually re-create all your edits and whatever, leading to highest costs for the artist!

Hope this all makes sense and if you need anything clarified, let me know.

I appreciate that I’m rather at the cutting edge of what people are doing with mastering (and might even trigger some people!), but this is the future and improving these tools as well as the workflow will help us contend with the onslaught (and maybe even kick the ass) of A.I. based, automated audio engineering!

Thanks!

1 Like
2

I understand and make note of the requests. I can see some of those making it either if a future patch (such as batch naming) or SL12 (module improvements).

2 Likes
3

I would like the instrument name to be listed first, to easily keep track of the instruments in the short name fields in SLP and DAWs.
For example: Bass - Let it Be.wav

4

I understand that and that’s actually how i manually do it. And that’s why i hate it when artists add their artist name to the beginning of a mix (totally unnecessary, like we keep all songs by all artists in the same folder!)

The only downside is if you want to keep the stems of various songs in the same folder, then you’ll see:

Bass - Let it be.wav
Bass - Hey Joe.wav
Bass - Billie Jean.wav
etc.

Which is a little more annoying when you want to find or delete the stems of one song or whatever.

To get over that, I’ll normally just add a random number to the start, like:
1 - bass - Let it Be.wav”

I love Total Commander for all the batch renaming tools, totally indispensable!

5

The way I see it implemented in SpectraLayers, is to let the user write its own pattern, such as
%projectname - %layer
with other possible fields such as %date, %time…

8 Likes
6

Nice, that’s a bit like how Samplitude does it:

samp
samp611×521 11.6 KB

7

I would like to have a folder with the song name for each song… :slight_smile:

8

Imo, the mistake in your request is the assumption of current underlying code capabilities circa end 2024.

Particularly in the request of the code to “hey, pull the kazoo-only out of this mixed mess” (and some mixes are indeed a mess :slight_smile: )

At the moment…imo…existing demix code is getting pretty good at pulling mixed drums (with or without artifacts) out of many overall mixes.

Which can then be split into further separated drum kit pieces as the code for THAT routine has a more bitesize set of audio-types that it has previously learned to listen for.

NOW…if you want the code to instead straight-away (circa December 2024) listen to, say, the below mix and…“hey just pull out the snare because that’s the only checkmark I clicked”…well…you’re suddenly throwing the code a whole pile of conglomerated sound that it hasn’t been trained to decipher.

Just my opinion of the state-of-code circa 2024 vs the state of 2028 code you’re asking for.

And then of course, the kazoo guys likely have to wait even longer :slight_smile:

1 Like
9

I can’t make any sense out of your post.

Am i asking that SL do anything that it can’t already do?

My point is about workflow.

Perhaps you listen to a mix and say: “i need individual access to just the snare, vocals and brass”. Then you want checkboxes for just those things.
SL already allows you to do this, but it’s a two-stage process. I get it, it’s because the drum unmix was bolted on to an earlier process. All i’m asking for is that it gets streamlined for the end user, i don’t care how it gets done under the hood.

Obviously if i give it a bunch of whale farts and ask it to extract the snare, it’s not going to do a great job, but that’s a bit of a straw man argument your putting forward there!

10

If I wanted to extract the snare from multiple songs I would have set up a Modules Chain (1. Unmix Song, 2. Unmix Drums (Drums)) and made a preset of it.
Then the workflow would be smoother afterwards.

1 Like
11

@ the devs, please consider everything i wrote in my first post, this comes from someone who is needing to do this many times and day and who loses masses of time on revisions because there is no automated way to swap out new mixes for old when you do the unmix thing.

Sure, if you only need to do it once a week, no issue, SL is fine like it is, but like i said before, please give us full time mastering people the tools so we can kick the asses of the A.I. systems!

12

Yes, I noticed :slight_smile:

1 Like
13

Hmm, the “Kazoo” layer; some things Ididn’t even realize were missing until I cam here to read this… :rofl:

14

Just time stamp it…how many “old” versions will you end up with?

In my time working in broadcast filenaming was an art form

date in reverse order being a perfect way to separate in the event your culture has the date and not the month first

personally I create a log for filenaming and copy>paste all my filenaming from my log make list…that way, if I make a syntax mistake, that mistake gets copied and can be easy to track down

that said I never use a search function in my OS, I always know where everything is…that’s just how I, personally, roll

the last thing I want is folder titled, say, “band name” and files inside said folder without the band name in the file name…I know it makes sense, yet easy to make a mistake

15

And if I understand the progress (generally speaking) there is more than 1 AI knowledge base (i.e. NN) used. I am guessing that the process for unmixing the band is a different network from unmixing the drums. Naturally, the code could run both steps behind the scenes, but that would potentially slow down the processing for folks who only wanted the unmix band part.

1 Like