As I work in both worlds ie mostly music and some film, I require a tool that covers them well. Spectralayers is best.
Nearly 10 years ago, I was handed a documentary funded by a charity and shot by ‘professionals’ who ruined the sound track beyond recovery. The producers spent $20,000 on a film they couldn’t use. Eg one scene had the narrator, who has a deep bassy voice, speaking right beside an idling diesel bus. The tools I had back then couldn’t handle the task.
I kept a copy of it believing there’d be advanced software one day that could fix it. Remarkably, SL10 was able to successfully unmix the “diesel track” and clean it to acceptable standards. In fact, that particular ruined sample has been quite handy helping me refine my SL technique as I learn optimal settings and plugin sequences for unmixing and repair.
I have great faith in Steinberg and eagerly look forward to SL developments.
Try not to view this comment as an opinion but more as a fact.
Both Spectralayers and Izotope are on the same level (in-terms of audio restoration and audio editing and spectral editing). Neither is more better than the other.
The only thing that Izotope has over Spectralayers is that (and it pains me to say this) it is more optimized. The gui(spectral resolution) is well optimized(even on low end hardware). I frequently adjust the resolution of the spectrum to edit audio and I find that each audio I edit has different degrees of magniscopes. Changing the resolution settings sometimes causes Spectralayers to lock up (especially when there are a lot of layers at a high resolution) as-oppossed-to Izotope where it seems to be a lot more optimized.
Other than that, Izotope is not better than Spectralayers and (ironically) Spectralayers gui is much more accurate. The infernocomposite view is actually the most accurate way to view audio and Izotope does have a similar view but I find that Spectralayers infernocomposite view is more easier on the eyes.
Here’s another little case study. I did a senior center performance yesterday with an 8-piece group. We record some of those, mainly for fun, but also to promote the band a little. I shot a video from across the room. I also do an up-close recording to replace the audio track in the video. I had planned to do a stereo recording using the XY capsules of a Zoom recorder, and to do a separate track for the vocal. When I arrived onsite, I realized I needed a TRS-to-XLR adapter to capture that vox channel, which I didn’t have with me, so I just went with the stereo recording. Also, on other occasions, I have tracked the bass and kick drum separately, but I didn’t want to do that much work – and we were not looking for studio-like quality.
Skipping to the end of the story, here is the video with improved audio track:
Here what I did:
I normalized that entire 60-minute stereo recording then ran it through the stand-alone SL unmix (just the band unmix, not the drums; if I had wanted more kick, I might have unmixed the drums too, but I didn’t care that much.) The unmix took about 90 minutes for the 60 minutes of material, but there are lots of football games on, so no problem.
I selected all the unmix channels except piano because we don’t use a piano. So I had vox, gtr, drums, bass, other and not-unmixed. I exported all those layers to WAVs and brought them into Cubase.
The unimixing wasn’t perfect, but it was good enough to allow some balancing and some different VST effects on the different channels. The only unmix layer that was significanty off was the vocal track. This band has 5 horns. Most of the horns ended up in “other,” but important fragments of the horns ended up in the vox track from time to time.
I needed to lower the drums. That was easy. I added a bit more verb to the drums than to the other material.
The bass track was OK. I used an Izotope plug to increase the transients and reduce the sustain to give it a bit more energy while keeping the ensemble sound light.
I needed to boost the vox because I didn’t have a direct capture of the microphone. No problem. I quickly used the Cubase alt-click to slice the areas where the vocal track was actually vocals. Then I used the clip-level gain to drop the level of the remaining material that SL had put into the vox track. This was a very quick edit. I didn’t try to put the loose material back into the track where it belonged. This procedure was fast and sounded decent.
So basically, other than the actual unmix processing, everything else was very quick. I probably spent less than an hour mixing the hour-long program, not counting the unmix processing time. And I am happy I could get this good of a result just from a single stereo pair.
Thanks for all these shares about unmixing etc. with SL10. Very helpful! I’ve been experimenting with my “Live” recordings made back in the 70s, and also some in the late 90s. The “late 90s” recording was more professional, where the sound engineer used more sophisticated mixing equipment at the venue, but I only have the stereo mix. This unmix sounds pretty good! However, the weak point seems to be keyboards… both in this and my other older recordings, and also I noticed on You Tube examples, the piano doesn’t come through as well. (There is a You Tube demo of SL10 which is of a live jazz song, sorry I cannot remember the link, but could send it later. In the live jazz demo, even though there is only “quiet” drums, bass, singer, and piano (no guitar), the piano still ends up not all on one track.). But besides all that, I do have a major gripe with the Isotope company. Their upgrade process is pretty frustrating, and not very “loyalty-friendly”. Even though I have spent around $740 over the last couple years on both RX Pro and Ozone Pro, Izotope expects me to pay around $199 per app to upgrade! And often, the upgrades only include a couple extra features, which I may or may not want. Then they try to upsell their “bundles” which include Neutron, Nectar, etc., etc., and the annoying part is often the same modules are included in several apps, which means that if you buy another of their apps, part of what you a paying for, you already have in their other app. This is the most frustrating part of Izotope, in my view. Ironically, I can purchase the SL10 cross grade for only $199.
Absolutely agree. I’d like to upgrade to Ozone 11. I have 10 through the production bundle. The “upgraded” bundle is mostly the exact product versions that I already paid for in the previous bundle. And if I just upgrade Ozone from the previous production bundle, that’s $150, which is ridiculous for that product.
Apropos to SpectraLayers, there are some nice features in Ozone 11. The one more relevant to this discussion is that with V11, any of the Ozone processing modules can apply either to the full mix or to one of the stems, which are separated in real time. That could certainly be useful in some mastering situations, but doesn’t offer any control over the separation process. I’d use it, but not for $150.
Likewise for Neutron. I have always liked that, but these days, the Cubase channel strip is so useful, I don’t very often have a strong need for Neutron.
I realize it is tough to make a buck in the music tech business, but I think Izotope was already on the way to pricing themselves out of the mainstream market, and that will only get worse now that they are coupled with Native Instruments. At those prices, they really require the customer to make a strong buy-in to the Izotope/NI ecosystem.
iZotope’s loyalty offers were always a bit idiosyncratic, but seem to have got even more so since the Native Instruments takeover. In particular, the iZotope website was not offering me an Everything Bundle upgrade, despite being a previous Everything Bundle purchaser as an upgrade from a single product suite (RX Post Production Suite IIRC).
In their recent Voucher sale, which I believe ended on 31 December, the e-mail I received included a link for Everything Bundle to which the voucher applied, so I was able to bring everything up-to-date for £155 - I was one version out of date on Music Production Suite, and 0.5 versions out of date on RX Post Production Suite. This was a much more palatable price than the website offer of upgrading just one of the suites.
I am getting more reluctant to upgrade iZotope software, especially in light of the continuing failure to fix the RX 10 DOP issues with Nuendo, the restriction of what ARA 2 support exists in RX 10 to MacOS with the only host officially supported being Logic Pro, and iZotope’s failure to release Dialogue Match for any host other than Pro Tools. Unless RX 11 works reliably with Nuendo DOP and has ARA 2 support for Nuendo on Windows, I will wait for a cheap deal before even contemplating an upgrade, rather than rushing to upgrade as soon as a new version of RX is released. It is about 15 months between my last paid upgrade and the cheap offer I took advantage of a few days ago.
I had some hopes that would get me there. But every combination I looked at, the voucher either couldn’t be used or didn’t actually save anything. That’s 45 minutes of my life I will never get back. Somebody at Izotope went to a whole lot of trouble to make sure the voucher would be of no use to a person who has had strong loyalty over the years. Maybe there is somewhere in the universe that this makes marketing sense, but I am not feeling it.
I truly have no idea what Native Instruments are doing.
I intended to ignore the sale, but I got a “Don’t miss your $50 Holiday Season voucher” e-mail on 28 December where there was an “Add to cart” button for Everything Bundle and I could apply the voucher code to the card for a discount.
I do not know why iZotope’s website loyalty offers page wouldn’t offer me a price to upgrade the Everything Bundle. Indeed, the whole loyalty offers setup is a mess - you’re left to work out which upgrades give you what rather than a clear statement: “upgrade this suite to the latest version and you get these new versions”.
Well, NI has always been about going their own way. It was really Steinberg’s claim to fame to lead in some of the early standards development. For a person composing in the DAW, especially composing techno stuff that relies on lots of unnatural sounds, the NI kit is strong. But you basically have to commit to living entirely in that ecosystem. And it ain’t cheap.
Are you using your CPU or GPU to unmix the songs?
If you have a decent graphics card the set it up to do all unmixing in SL settings you can cut down all that to seconds.
Unmixing 6:33 song took me less than 30 seconds for full song unmix on Best settings.
Just a heads up to those that might be unaware of such option.
Where is this set? In the stand-alone app, Preferences → System → AI Processing Device only offers “CPU”. If that is where it is set, that probably means it doesn’t support my graphics engine, which is pretty old by now.
I have been putting off replacing my main studio system for several years, because it still works well and is pretty fast. It is probably time to replace the whole system in 2024.
As an aside, I worked in system engineering for a large computer company much of my professional career. In the late 1980s into the 1990s, we saw a great deal of enthusiasm for “parallel computing”. This took many different forms. Initially the concentration was on “multiprocessors” (equivalent to the development of multiple cores on the earlier Intel chips) where the OS would dispatch completely different workloads (different apps) to run on the individual processors. But as the 1990s progressed, the scientific computing (i.e. “supercomputer”) people became more interested in large scale parallel processing where there might be dozens (or eventually thousands) of processors available to take on pieces of the work. This level of compute power was necessary for many meteorological and geological applications.
The problem was that dispatching completely different applications is relatively easy, but using large-scale parallel processing required developing algorithms that could attack a computing problem in many sections simultaneously. That is not easy, and while hypesters were always coming up with hardware systems that had lots of processing elements, the software was always way behind that.
The most successful formula for large-scale parallelism is when it can be built into an architecture that is mostly transparent to the programmer. For example, parallel database technology has become enormously successful because the nature of relational databases is that many of the operations can be performed on sets that can be processed in chunks by separate processors. The programmer enters SQL commands, mostly unaware of how they will be decomposed and processed.
We don’t see VSTs, in general, operating in parallel (within the function of a single VST), but some DAWs are getting good at using many (typically 10 or fewer) cores when there are many VSTs in use.
Crypto mining laid the foundation for really large scale parallel processing using the GPU architecture that can provide thousands of processing elements. It seems to me that the AI world is following that path, maybe even using some of the same underlying software layers, to give us a high degree of parallelism for the AI part of the problem.
All of this is to say that in 2023-2024, we are just now actually realizing some of the dreams people were pitching on a daily basis back in 1994. And what they were trying to sell for $10,000,000 in 1992 is now becoming possible for $2000.
What is “inferno composite” view? Also, in RX, I can set the “View” to “Show Channels” as one (mono) channel (composite view), which really helps in viewing the entire frequency range of the music. Is there a similar setting in SL10?
It is the most accurate way of spectral analysis and spectral editing. Put it this way (a better way of reiterating the scenario), most spectrums/spectrometers/spectrographs/spectrograms more-or-so gives you a 2 dimensional analysis visual of your data, for example in audio spectrums it is usually Volume/Amplitude X Time (X over Y). In video analysis(for example color grading or shader editing), it is the same concept. The composite views view is exactly what it means. In composite view you get a more 3 dimensional aspect to your analysis. For example instead of Volume/Amplitude X Time, now its Volume/Amplitude X Time X dynamics. For further example it is X by Y by Z(Zed being for Dynamics).
If you mean by adding the sum of both channels to equal the output view of both channels while viewing only one channel (meaning taking the visual view of both right and left to add together as one channel while you can still hear both right and left audio while only viewing one separate channel as in mono visually), then as far as I know only FL studio does this. If you load any file into Edison it will play as a normal file(as stereo) while visually showing you one channel(giving you the impression it is mono when in reality both of the sums of those channels are added visually).
If you mean that the stereo and mono channels are not as intuitive as dealing with layers (as far as solo-ing and muting goes) then the main developer is already aware of that.
What I mean is: RX can display the spectrogram as one combined channel rather than left channel and right channel. This helps because on your screen, you can see the entire frequency range from 20 HZ to 20 KHZ.
Actually, I do see that I can set the view to LEFT channel, and the display will be full-screen, and if I select a frequency, it will be selected in both channels. So that is a work-around.
