SpectraLayers 10 is a life saver, for legacy projects

[mhschmieder]

I finally started playing with SpectraLayers 10 last night, a couple of months after buying it, and it blows iZotope RX out of the water when it comes to unmixing legacy tracks. Almost no artifacts afterwards, and it doesn't even have parameters like RX, for most of the tasks! A few more choices for what to unmix to, and what happens with non-unmixed elements.

Additionally, I find the workflow and overall interface of SpectraLayers vastly superior and more flexible and intuitive than RX, even though the latter has been part of my weekly workflow ever since its initial release.

I have some recordings from a 1980's project that got uploaded to digital from cassette back when most of us only had two-track options in the digital realm (pre-computer recording days), meaning everything is two-track stereo mixes, sadly.

I separated the left and right tracks and unmixed them independently, as the quick cassette mixes that prepared for going from four tracks to two, had some separation that way, except for the earlier bounces done to make room for overdubs.

The Irish bodhran features on these tracks vs. a drum set, but combining the non-unmix track with the drum track yields the full bodhran sound, with no artifacts. My bass was hard-panned left for the digital bounce so it's easy to separate.

The only drawback, also true of RX, is that harmony vocals cannot be unmixed, unless it is a spoken word project. As I hard-panned the harmony vocals to the right before the 1990's digital transfer, I can probably subtract the left track to get it.

So far I have only found one thing that didn't work as well as I hoped, but as my CPU can't handle real-time monitoring, I was too lazy to try different percentages of de-reverb on the bass track, which I think my bandmate effected on the mixer.

I will simply use SPL's de-reverb plug-in once I import the unmixed bass track to Digital Performer.

Next up, I will try the audio repair functions of SpectraLayers vs. those of RX Advanced.

[monkey man]

Sounds kinda-exciting Brother Mark.

Awesome mate.

[stubbsonic]

Yesterday, I did a test project with the trial version of Spectralayers 10.

We needed to tune some vocals that were recorded at the same time as acoustic guitar. The guitar was recorded in stereo. So three tracks Gtr L, Gtr R, and Vocal. Of course, all three had bleed.

So after unmixing in SL10, each track now had three tracks: vocal, gtr, and "non-unmixed" (i.e., whatever AI couldn't parse).

At first, I thought the result might not be useable, because I only listened to the "goal" tracks. However, when I played all nine tracks, it sounded like the originals-- the artifacted gating thing was no longer noticeable.

Next step, tune some vocals in DP, job done. Next, turn back on all the tracks again-- and viola. The artist was pleased and relieved. I still need to hear back from the project's producer, but it seems like the test was successful.

I'm hoping they add new instrument classes to their AI picks for unmixing. Organ would be an obvious one. Since they have one for drums, perhaps after other types are unmixed, then they could have a second step to unmix parts.

It is possible that Spectralayers trial gambit paid off. The idea of being able to pull out instruments in this way seems quite handy. [EDIT] Oh, I didn't realize I was looking at the competitive upgrade price-- MH corrected my misunderstanding.

I have no experience with Melodyne's DNA (direct note access) thing. I wonder how it compares.

[mikehalloran]

Some of the advertised features of SpectraLayers are found Ozone Advanced instead of RX Advanced, both of which I have. At first glance, I'm not seeing what else SpectraLayers brings to the table but, at $199 for the competitive upgrade, I'm not all that sure I care.

I do have an old 2-track cassette master on which to test everything. My wife lost the 4 track masters a decade ago so this will be interesting.

[mhschmieder]

Mike, I have not found RX's music rebalance tool to do a good separation job, but it did help me rebalance, which after all is its name. SpectraLayers is the first tool I've used (of many!) that actually succeeded n letting me create usable artifact-free stems from legacy mixes. It will be interesting to hear your experience.

[mhschmieder]

https://www.soundonsound.com/reviews/st ... ers-pro-10

[joelfriedman]

Wow. That's impressive.

[mhschmieder]

I'm keeping iZotope RX Advanced for now, and may upgrade to v11 in May, as I think there are still some things it does better, but my original post was primarily about the unmix feature in SpectraLayers, which in my experience blows Music Rebalance out of the water, at least in that role. The RX tool is still very useful for actual rebalance vs. stem separation though.

What I think I am finding so far, is that most of the audio repair tools in RX are a bit further along than those in SpectraLayers, but I think the de-ess tool in SpectraLayers works a bit better.

The one thing I haven't figured out in SpectraLayers yet, is how to use the de-bleed sidechain to effectively unmask faint background instruments that aren't strong enough in the miked source to be good candidates for an unmix tool.

It's probably just a steep learning curve on terminology and workflow, but I gave up due to a deadline and used the incredible new pure:unmix tool from Sonible instead, which didn't fully remove the bleed but that's not its primary intent anyway; it simply adapted itself very well to that workflow and goal.

What I thought I could do in SpectraLayers is to pull in the prime track for the bleed source, which in this case is the bass guitar mic track that bled into the drum overheads, but I haven't figured out how to do that as it rejects every attempt I make to select a source.

[tremo]

I have some recordings from a 1980's project that got uploaded to digital from cassette back when most of us only had two-track options in the digital realm (pre-computer recording days), meaning everything is two-track stereo mixes, sadly.

I'm working on a similar project right now, trying to clean up (and possibly enhance) songs my band recorded in the early 80s. All analog, and all recorded in professional studios, but sadly I only have cassette dubs. So I compared trial vesions of SpectraLayers and another unmixer, RipX DAW Pro. Results were pretty close, but I ended up buying RipX. I just worked better for my particular case (YMMV).

https://hitnmix.com/ripx-daw-pro/

One thing I like is that by default RipX generates more individual stems (or "Layers" as RipX calls them) than SpectraLayers. You get separate Kick, Drum and Percussion stems, for example (although I always just merged the perc and drum layers). We had a couple songs with guitar and alto sax, and RipX did a good job of separating them -- a few sax notes ended up on the guitar layer, but again, you can cut/paste stuff from one layer to another. Vocals were separated quite cleanly.

There are still artifacts audible when you listen to the separate stems, but when you bring all the faders up in a new mix, everything merges together pretty well.

I guess the dream is that someday Peter Jackson will release the magical AI his team used on "Get Back" (and which Bob Clearmountain used to do wonderful Dolby Atmos mixes of the first two albums by The Band)as a commercial product...

https://youtu.be/DbJf59h6Fe0?si=EZLc7bE5OpOAlgnj

[mhschmieder]

Interesting, RipX is not one that I had yet heard mentioned anywhere else!

[stubbsonic]

RipX Pro is on "spring sale" for $158 USD (reg. $198).

I'd be curious to know how well it compares in separation quality & features to Spectralayers, since RipX is 2/3 the price (currently nearly 1/2 the price).

[mikehalloran]

Interesting, RipX is not one that I had yet heard mentioned anywhere else!

RipX is big in hip-hop which is all I know about it.

Otherwise, I've been playing with SpectraLayers Pro 10, Simple Stems 2, the Stem Focus (Ozone 11 & Advanced), Master Rebalance (Ozone 10) and Music Rebalance (RX 10 & Advanced) tools to save time in a project I'm working on now. Except for SpectraLayers, all are based on Spleeter freeware. BTW, despite the nonsense posted on the internet, Oz 10's Master Rebalance is a tinkertoy compared to the versions found in Oz11, RX10 and SimpleStems 2.

All managed to remove a bass track I needed gone from an edited mix and none could isolate a claves track. Fortunately, any dynamic EQ could zap the claves and that's what I did.


Simple Stems 2 is the easiest to use and I don't regret the $15 license I paid last year. I could pull it up as a plug-in in DP but can't make it work. The stand-alone app does what I need
https://www.stagecraftsoftware.com/prod ... mplestems/
He doesn't call it shareware—but it is, being fully functional on download. There are discount codes all over the internet though I'm not finding the $15 code I used last year at the moment. If, however, you have Ozone 11 or RX 10, you don't need it.


Back to SpectraLayers Pro 10. I have to remaster a cassette album that I did in 1991 and will probably use it for this. Patting myself on the back for having bought it for the $199 intro price a couple months ago. I haven't figured out how to use it as a plug-in in DP though it did Verify and I have it enabled—again, there's the stand-alone app. Oh well, there's a 104 page User Guide that I have to read, I suppose, and YouTube videos to watch.

[HCMarkus]

I'm just gonna' try to mix it well the first time around.

[labman]

I'm just gonna' try to mix it well the first time around.

[mhschmieder]

I'm just gonna' try to mix it well the first time around.

Famous last words.

Of course, most of us are dealing with projects we did not control, kind of like how most software jobs are cleaning up the last person's mess.

The other main category is premature digital transfers that forced us to do track reduction from the original analog, sometimes to stereo (mixed, not even dual mono!), followed by the loss of the multitracks and/or the gear that could play them.