• Hi and welcome to the Studio One User Forum!

    Please note that this is an independent, user-driven forum and is not endorsed by, affiliated with, or maintained by PreSonus. Learn more in the Welcome thread!

Stem separation

"I did a side by side comparison with Studio One 7 and Logic Pro wins. I really would like to see Studio One catch up."
"I would just like to see more speed here from Presonus."
"I don't expect perfection, I expect progress."


This technology feature is very new in DAWs. There are not endless basic technologies in the background. One uses this, the other uses that. These may be business decisions and strategies that are not immediately apparent to the user. There is progress all the time.

The important thing is that there is not THE ONE algorithm that simply does everything perfectly for all use cases. So you can't say that this is better than that. Because that only applies to this one test. And only according to the interpretation of a user of this one test. But you can't take this approach when developing a DAW for millions of users with millions of different music styles and use cases. That wouldn't be far-sighted for a DAW.

Just like the call for more speed from PreSonus. Studio One in particular has a very tight update cycle. Chasing after a single technology doesn't do justice to all the other users who don't care. Remember back when DAWs started modeling guitar amps, which was really a special task for 3rd party plug-ins. Today, almost every recording DAW has such a simulation. And you can work well with all of them. That's kind of how you have to imagine it now.

Best
Ari
 
Hi Ari,

Thank you for your contribution to this thread and your perspective on the subject.

As one of the authors of the quotes, I’d like to clarify my point again:

1) I associate progress and speed with the ability to release updated models (and the 1.8GB download is nothing more than that) outside of a release cycle, specifically when they’re finished. Users who aren’t interested can safely ignore this; they probably haven’t installed Stem Separation anyway.

2. I believe point 1 is crucial because, based on my tests, the model update along with the 7.2 release hardly brought any significant improvements. While the topic is currently labeled as “work-in-progress” everywhere, one would expect at least audible improvements compared to the previous version, such as fewer artifacts.

Overall, Stem Separation in Studio One performs well, and of course, it can’t be compared to specialized products developed exclusively for that purpose.

In conclusion, I would like to reiterate that if you have already cleverly designed it as a separate download, you are more than welcome to use it accordingly and more frequently!
 
Last edited:
"based on my tests"
That's what I was trying to say before. ;)

"it can’t be compared to specialized products developed exclusively for that purpose."
The Stem Seperation is a product exclusively developed for this purpose. As I said, there are not unlimited of different "models" out there. All of them uses a similar base. This has nothing to do especially with Studio One or any other DAW. And if there are updates, I am sure you will get them.
 
I used Stem Separation yesterday to remove the drums from a vocal track .It was a jam with friends...just three tracks In Studio One I extracted the drums... reversed the Phase in the Mixer Played the vocal track with that and moved the fader of the (reversed) drums until the drums disappeared Did a mixdown of the two tracks and brought it back into the project Crude but it sort of worked
 
"reversed the Phase in the Mixer"

🫣 Sorry but you reversed the polarity. 🫢
Precision is key.
 
"The important thing is that there is not THE ONE algorithm that simply does everything perfectly for all use cases. So you can't say that this is better than that. Because that only applies to this one test."

With all due respect, I can say for certain that the Stem Separation in Logic Pro give's me better result with all the vast variety of audio files and styles of music I have worked with and I have done some extensive work in both DAW's. Have you had a chance to do some comparison for your self? Let me be clear, I'm not saying Studio One's algorithm is bad, as a matter of fact I give it high praise for what it is. I'm just being honest when I say Logic Pro is winning in my book. At least for the time being. We know that could change as things progress. The AI technology is growing and changing with every passing day. I really hope it does because I have high hopes for Studio One in this area. It's easy to say all things are subjective to this or that but the proofs in the results and that's for each user to decide as I have. I can honestly say before the Logic Pro 11.2 update, from my experience, they were pretty much on par but now, there is a clear noticeable difference in quality as well as the added Stems in Logic Pro. As to whether the Stem Separation feature is of some level of importance to PreSonus to continue it's progress over other features is still yet to be seen. I sure hope so.
 
Last edited:
It was not my intention to emphasize one or the other algorithm. In fact, I did the exact opposite. If another product currently works better for you for a task, then use it. It's that simple. It's basically the same discussion as with DAWs per se or even plug-ins. They're all just tools. Use what you can and want to work with. The main thing is that you get the result you want.
 
Bearing in mind, every source recording will be different. Spectral Layers Pro can analyze differences, but frequencies unless quite varied, won't always reveal unique patterns.
Two guitars or two anything will likely always yield different results. Especially within some busy stereo field where similar EQ is going in between them.

Definately try before you buy. I was able to use Studio One's stem separator in a recent song of mine called "Rio Amazonas". I had this ending where a Toucan was squawking, and unfortunately a cameraman was firing off pictures and a woman talking. So between the shutter noise and the voice, I couldn't use the video and audio segment. Enter the stem separator. I extracted the audio, then had to re EQ and invert phase, and then it extracted the noises perfectly. Things just always don't work in audio and will require some shifting and coaxing, but I'm really glad Studio One now has the feature.
Sometime, you have to re.eee.eeech for the candy bar. 😉
 
Stem separation is so situational. I'm curious how people use it.

Recently, I used it to extract the vocals from an album's final mixes (already normalized to the same LUFS value) to check for consistent vocal levels. When you listen to an album, by the time you reach song #5, there's no way to remember what level the vocal was in song #1 but you might think it was a little bit softer or louder than other vocals. Extracting and comparing stems was an easy diagnostic tool, especially with the Song/Project page synergy.

In this case, stem separation had nothing to do with audio quality, and the stems did not appear anywhere. But stem separation avoided having to jump around songs on the Project page timeline and second-guess myself about levels. There is no doubt that editing the level on a couple vocals made the overall album sound more cohesive.

Of course, you still have to listen to confirm that any edits created the desired result. Numbers are guides, not gods. But they sure are helpful :cool:
 
Craig Anderton said:
Stem separation is so situational. I'm curious how people use it.

Recently, I used it to extract the vocals from an album's final mixes (already normalized to the same LUFS value) to check for consistent vocal levels. When you listen to an album, by the time you reach song #5, there's no way to remember what level the vocal was in song #1 but you might think it was a little bit softer or louder than other vocals. Extracting and comparing stems was an easy diagnostic tool, especially with the Song/Project page synergy.

In this case, stem separation had nothing to do with audio quality, and the stems did not appear anywhere. But stem separation avoided having to jump around songs on the Project page timeline and second-guess myself about levels. There is no doubt that editing the level on a couple vocals made the overall album sound more cohesive.

Of course, you still have to listen to confirm that any edits created the desired result. Numbers are guides, not gods. But they sure are helpful :cool:
Click to expand...
Used it yesterday to remove vocals from an acoustic guitar microphone on an album I am mixing. I wanted to push the lows and the low mids a bit and kept finding the vocals in that microphone were messing with my reverb. I removed the vocals and was able to push the low mids cleanly and had no phasey vocals in my reverb.
 
By the way, I use it to separate drums from real existing songs. Why? Because I work a lot on room simulations (algorithmic or with impulse responses) and need good testing stuff all the time. Vocals are pretty interesting too of course for this purpose.
 
I'm an occasional user of stem separation. Sometimes to "unmix" old songs of mine where I've long since lost any written notes about it. You can separate it out, stuff the melody(s) into Melodyne to get a fairly decent MIDI snippet and chop the drums up so they can be replaced with other sounds. I also occasionally used it to nick the vocals from a commercial song to cut up for samples. I don't publish any of my stuff so sample clearance is not an issue. Just a bit of fun really.

Incidentally, Steinberg's SpectraLayers 12 is just out and it is supposed to have taken stem separation to a whole new level. It will be a while before I get my hands on it but I will report back if this thread is still alive.
 
Craig Anderton said:
Recently, I used it to extract the vocals from an album's final mixes (already normalized to the same LUFS value) to check for consistent vocal levels.
Click to expand...

If you're interested in finding out more about this technique, it's described in this week's installment of my mostly-weekly PreSonus blog post.
 
{EDIT] Per Forum Terms and rules the video link was removed (contains profanity).


This work was done with an Audio CD.

In previous audio separation processes, the source file was in the Video Clip mp4 format (160kbps AAC), which made mixing a bit problematic.

This time, the results were very successful with the INXS - Need you tonight (32bit 48khz) lossless source file. We specifically tested S1 Separate Stems against SpectraLayers 12. The results were better for drums with SL12. The bass and melody channel effects are similar. I'd say it's the highest quality separation process we've ever done. Pay particular attention to drums and vocals.

Studio One audio separation (the new Stem Separation version 1.1.1.106) was used, and this updated version is better. However, the SL12 audio separation, especially the drum quality, is impressive, and the vocals are very clean with v12. The source audio format and quality are important factors.

Bonus: Short film for you (from Anna 2019 Luc Besson)
 
imacken said:
Thanks but removing drums and/or bass is easy, and S1 can do that. My need is to separate lead and rythym guitars, which S1 cannot do.
As recommended above, SpectraLayers Pro can do this by analysing and making selections on the spectrum of the separated guitar track. I have a thread going on this in the forum mentioned above.
Unfortunatley, the Pro version of SpectraLayers is around £250, which is way more than I could possibly justify.
Click to expand...
Hi, if you don't want to spend too much on stem separation, you can try the La La Lai website.
I had to do some testing on a song that SO wasn't playing well, and I have to say that La La Lai's algorithm is really efficient.
Regards
 
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

This audio file was separated using mp3 format and 160kps. We tested it separately with Studio One v7.2 and SpectraLayer v12. When the music file data is insufficient and in low format, separation with both plugins produces similar results.

Drums: Pro-Q4, Pro-C2, CLA Vocals, and SPL Transient were applied.
Vocals: Pro-Q4, Pro-DS were applied. Waves H-Delay and Compressor were sidechained with the FX channel.
Other channels: Pro-Q4 transients and Soothe2 were added.
Extra: A rhythm was added to the darbuka (hand drum)

Bonus: Short film for you (from Drop 2025 Christopher Landon)

Important: LED monitors, be quiet!
 
You must log in or register to reply here.
Back
Top