If you're mixing live vocals YOU NEED TO READ THIS!

Well I must say I’m beyond blown away by SL10 Unmix Song feature! I’ve been using SL 8.0.10 to unmix live vocals from the mic bleed (mostly cybmbal and drum bleed) and it’s done a pretty decent job up until now. I had an issue when it was updated to 8.0.20 that made the process take 7-10 times longer. Between SL8.0.20 and on to v10, the speed difference is not that great, but the QUALITY of the audio is UNBELIEVABLE! Here’s my test results:

24/48 wav file - 4:48 minutes
SL 8.0.10 - 0:13 render
SL 8.0.20 - 1:35 render
SL 10.0.2 _ fast mode - 1:40 render
SL 10.0.2_Best - 5:00 render

So the times are atrocious comparatively, but the difference in separation is what I’m really impressed with. That being said, I compared Fast mode to Best quality mode (SL10) and again, the difference was small (quality of result). There was definitely a noticeable difference for the discerning ear so if you’re wanting the BEST quality result, “BEST” is worth the extra time. :slight_smile:

I didn’t notice as much of a difference between Fast and “Regular” however. But all material is different so your mileage may vary…

Like I said, I use this feature to de-bleed live vocal mics (which this feature excels at). I do LOTS of live recordings and mixes and running 8.0.10 on a 2 hour + file takes about 5-8 minutes. In SL 8.0.20 and later this takes about 45mins to an hour! If I’m de-bleeding 4 files, you see where the time saved is HUGE.
Anyone interested in this for de-bleeding live vocal mics, I just uncheck everything except vocals and then mute that channel and export.

Running this process in SL8 vs SL9, the quality difference is negligible. When I say that, i’ve done a comparison by importing both to the DAW, flipping the phase on 1 track, and the meter shows the difference at -160db. That’s really, for all intents and purposes, no difference at all. None that you’d ever hear in a mix, no matter how sparse.
The test I conducted was on pretty much a worst case scenario. Client sent me live recordings where the vocal mics were about 6-8 feet in front of the drums. Closed room. When I ran it in SL8 and completely dumped everything but the vocals, there were still some drums in the background when the singer was singing. Enough to make an audible difference.
Running it in SL10, there were NO DRUMS LEFT! None! I mean they were non-existent behind the vocal. I’ve tried a lot of tools (RX10, etc.) and nothing even comes close to this kind of separation. The artifacts were very minimal as well. Acceptable when you put them back into the mix.

Kudos to the Spectralayers team for this upgrade in quality! I will be upgrading this ASAP and keeping SL8.0.10 for when I need something fast, but SL10 is my new goto.
Secret weapon just got an upgrade! Shhh

4 Likes

Astonishing. I’ll buy it the second it’s on sale.

Do you qualify for a crossgrade? If so, that’s probably more than any normal sale will be.

1 Like

Sadly not. I only have the version that comes with Pro, so not eligible.

I hope I am posting this in the appropriate place. I am testing the demo version of Spectralayers 10. I am working with some old, live recordings, which is why this question goes out to others who may be using this for the same situation. I am just wondering, do you have good success with trying to unmix live recordings that were mixed as just stereo? Because in my quick test, I am having problems. (1) The biggest problem is when I unmix my live songs, I cannot get a decent bass track. In these live sessions, I suspect this was on a stage where the (electric) bass was not definitely not direct miked, and probably did not even have a mike directly on it. (2) The electric piano track come through only sometimes (loud spots). A little bit of piano comes through on the “unmixed” channel, but not reliably enough to try merging. (3) Remarkably, the saxophone comes through on the “other” channel, but unfortunately it also bleeds into some of the vocal and guitar tracks as well (but not all the time). (4) The test I did for this was in BEST mode unmix. Does this mean that SpectraLayers is more effective on studio-quality mixes? Thanks!

1 Like

I have had the same situation. Even though the human ear can discern teh bass line – or at least most of it, SL may not get much (or any) of it. It is hit or miss. IN some cases, I got pretty decent results if the bass was just a little higher in the mix. There are some suggestions here: SL10 unmix, no bass extraction - #20 by dibellodesign

But mostly, I suspect you will have to wait for an improved unmix version for some of these. Anything you can do to get more of the bass into the original recording will help of course.

Not sure about unmixing live stereo recordings. My post was very specifically about bleed in a live mic that’s recorded multi track. Have you demoed iZotope RX or Acoustica? Not sure if they would be any better for this, but for live UNMIXED vocals, this de-bleeds better than anything else I’ve tried to date.

Yes, I own RX10. TO be fair, RX10 is not that great with bass tracks either. I tried the same song with RX10 and the bass line came out better than SL, but still not acceptable. The other problem is with unmixing piano. I notice, even in You Tube demos, folks are having trouble getting a decent piano stem. SL seems to pull out only the loudest parts of the piano track; then you have to go to the “Other” or “non-mixed” stem and cut and paste the portions of piano that ended up there. I wonder if SL’s algorhythms work better with digitally recorded music? Because the two tracks I have tried with it, so far, are old original analog recordings.

1 Like

I am sure SL can be improved to do better with bass. But we should all appreciate that there are limits to what you can get out of unmixing. Unmixing is magic, and it may be the most important tool in the past 5-10 years for people recording live performances.

However, the fact is that when you bounce sounds all over a room and then put it into a stereo pair, much of the original content is lost due to phase cancellations on stage. Even with the best possible unmixing technology, we are going to lose much of the content compared to the case where we might have captured each source to a separate channel (even with some bleed.)

And on the positive side of the ledger, SL does a tremendous job on drums. It is hard for me to imagine SL or any other tool getting much better on drums. For me, with any live recording short of tracking every instrument separately, almost always the drums are too loud in the mics. Being able to remove this or cut it back can make a huge difference.

For me, I can get a pretty decent result with an 18-piece live big band putting a stereo pair in front of the horns, running a DI line from the bass, taking a split of the vocal mic(s) and putting an area mic near the grand piano and guitar. (i.e. 5 or 6 tracks)

1 Like

@ErnieMuse

Please read this, it should not only help but enlighten on the matter. Also would encourage you to read the rest of topic

There’s no limits. I’ve been able to solve every audio problem I’ve ever had using Spectralayers. The only minor issue is (unlike video editing vice-versa) there is no real-time editing of audio for video(meaning seeing the video playback in real-time as you’re editing the audio).

What? What does phase cancelation have to do with audio recording quality?

1 Like

On a live stage, there is cancelation happening everywhere, all the time. If you run the sound into a stereo pair, you will get only the resulting sound from that exact location after all the cancelations have happened. Therefore, when you unmix, there is a certain amount of the content that is simply not recoverable because it was never recorded.

A simple example of this is if you have two trumpet players sitting side by side, playing the “same” pitch, but one player is at A=442 and the other is A=440, you will get an obvious pulsation. If you are recording that combined sound to a single channel, that recording will include the pulsation of energy caused by the phasing cancelation. That lost information is never recoverable because it was never recorded. If instead, you had recorded the two players separately into two tracks, you would have all the information. Of course, if you mix them later, that same phase cancelation will occur if all other factors are identical between the two channels. That example is not an “unmix” scenario, per se, but a simple illustration of the problem.

The upshot is that a live stereo mix will have less information than if all the instruments were captured individually. Therefore, when unmixing a crowded stereo mix, you should expect the stems to sound thin, and that’s actually pretty obvious in all the unmixes I have done, regardless of whether it is SL or Ozone/RX.

But if I can solve an otherwise fatal problem by using the unmix tools, I can live with a little bit of thinness in the process.

I don’t think I am saying anything very profound here. If I track 10 channels, I am recording about 80 MB per minute. If I just record the stereo from some arbitrary position in the room, I am capturing about 16 MB per minute. Of course, there is less information available in the latter case.

1 Like

@cparmerlee

Okay! Maybe I got the terminology mixed up. I couldve sworn that what you are describing is the doppler effect not phase cancelation.

Thanks, I appreciate this! But as I started to read all the various posts, I was realizing that it may take me less time to re-record the bass lines than to do all this trial-and-error processing. (1) Some poster said that their unmix put the bass line in the guitar track, that same thing seems to be happening to me. In fact, I was trying to think whether the guitar player and bass player were actually sharing the same amp on that gig!

Perhaps we use different terms. I generally think of the Doppler effect being caused by either the source or the destination of the sound being in motion. That doesn’t cause a loss of fidelity per se, only a pitch shift, because of an increased or decreased arrival rate of the waves at the observation point (microphone).

That would be very interesting. If they were both going through the same guitar amp, it is possible that the bass inherited a profile that did appear more like a guitar than a bass. Of course, players have been moving to extra strings on the bass (6 strings) and electric guitars (7 strings) in order to allow more overlap between the two instruments, and this will probably give all of the unmix products major fits.

1 Like

Not sure if I am posting in the correct category for this question, but: is there a comparison of RX Pro vs SpectraLayers 10 here? I’d be very interested to get viewpoints. I have owned RX Pro for a couple years. I like it, but looking back, I have spent $641 including the various upgrades, and I haven’t upgraded past RX9 because I couldn’t justify paying more money for the new features. I see that Steinberg offers a SpectraLayers 10 cross grade for $199; but on the other hand, I cannot tell if SL can do all the same processes as RXPro. I am looking at the Demo of SL and, since the interface is slightly different, I really cannot quickly compare the two. One thing I do like about RX is that there are many presets for the various tools, so that you don’t have to start from scratch every time you need to perform an operation. Thanks!

1 Like

Have you found way of fattening those thin, unmixed tracks using Steinberg or other tools?

I’ve also used both and prefer Spectralayers overall. RX10 is aimed more at film and TV production. According to Bard:

RX10 vs. Spectralayers: Audio Editing Powerhouses Compared

Both RX10 and Spectralayers are industry-leading audio editing software programs, each with its own strengths and weaknesses. Choosing the right one for you depends on your specific needs and workflow. Here’s a breakdown of their key differences:

Focus:

  • RX10: Primarily excels in audio restoration and repair, offering a vast array of tools for noise reduction, clipping restoration, spectral editing, and more. RX10’s strength lies in its automation and batch processing capabilities, making it ideal for repetitive tasks and large workloads.
  • Spectralayers: Specializes in spectral editing and source separation, providing unparalleled precision and control over individual audio components. Its visual spectrogram interface allows for highly detailed manipulation of audio, making it perfect for tasks like vocal isolation, de-essing, and harmonic editing.

Ease of use:

  • RX10: Features a streamlined interface with clearly labeled tools and functions, making it accessible to users of all skill levels. Its modular workflow allows users to chain specific processes together for efficient editing.
  • Spectralayers: Has a steeper learning curve due to its complex spectral editing interface. However, its powerful tools offer immense flexibility and customization for experienced users.

Features:

  • RX10: Boasts a wider range of features, including:

  • Extensive noise reduction tools: Hum removal, click removal, mouth clicks, vinyl crackle, etc.

  • Spectral editing tools: Repairing clipping, removing pops, and editing transients.

  • Music rebalance and unmixing: Isolating specific instruments or vocals.

  • Dialogue editing and repair: Removing plosives, sibilance, and other unwanted sounds.

  • Batch processing: Automating repetitive tasks across multiple files.

  • Spectralayers: Focuses on spectral editing and source separation, with features like:

  • High-precision spectral selection: Isolating and manipulating individual audio components with pinpoint accuracy.

  • Advanced source separation: Extracting vocals, drums, and other instruments from mixes.

  • Harmonic editing: Fine-tuning the pitch and timbre of individual sounds.

  • Spectral painting and resynthesis: Creating entirely new sounds from scratch.

Overall:

  • RX10: The go-to choice for comprehensive audio restoration and repair, with user-friendly tools and batch processing power. Ideal for audio engineers, podcasters, and anyone dealing with noisy or damaged audio.
  • Spectralayers: The ultimate solution for advanced spectral editing and source separation, offering unmatched precision and customization. Best suited for experienced audio professionals and sound designers seeking in-depth audio manipulation.

Ultimately, the best choice for you depends on your specific needs and budget. If you prioritize ease of use and versatility, RX10 might be the better fit. If you require maximum control and precision for spectral editing and source separation, Spectralayers is the clear winner.

So far, I have used most of the unmix layers in the reconstructed mix, so they don’t sound particularly thin. I usually mic fairly close to the ensemble, so the recording is rather dry. Adding some modest reverb adds fullness. If you have taken vocals off to a layer, adding some tasteful about of chorus can give the vocal track more presence. I also use Izotope Ozone, which has several exciters that can give the sound a richer quality.

1 Like

I don’t know that the functions are that much different. Certainly, Izotope has priced the Advanced version of RX way beyond what most hobbyists would want to spend. The Advanced product is clearly targeted at commercial operations (TV stations, movie studios etc) that have no trouble paying the price.

And you are right, most of the features that are exclusive to the Advanced level are things I would rarely use in my work. There is a concentration on dialog and ambiance matching – things more important on a movie production, or things that a TV reporter might deal with a lot more.

1 Like