Reduce/Remove Reverb from 2-Track Mixes

[Gutter Puppy]

Hi -
Is there anything out there that is effective at reducing/removing reverb from an established/final mix (eg. with multiple instruments + vocals) without screwing up the overall tone/sound of the mix ?

Thanks !!!

[adrian_k]

I think Ultimate Vocal Remover will do this, although I haven’t used it for that particular task. It’s free so nothing to lose.

A quick internet search should clarify…

[James Perrett]

As Adrian says, you can do this with the Ultimate Vocal Remover but you probably want to split the mix into separate stems first as different instruments will probably require different amounts of reverb reduction. Go to

https://ultimatevocalremover.com/

for the main software and then add the UVR-DeReverb-aufr33-jarredou_4band_v4_ms_fullband algorithm from the VR Architecture in the Process Method dropdown.

I haven't been following developments closely over the last few months so there may be more recent algorithms available but my suggested workflow would be to use the Demucs V4 algorithm to split the mix into stems and then run the DeReverb algorithm on each stem. I tend to use the default settings but it may be worth experimenting.

https://music.youtube.com/playlist?list ... VrqsTjYcSI

Is an example of something that I split into stems, then removed the reverb, remixed and then added a little different reverb in order to make things gel together better.

[jimjazzdad]

I haven't had occasion to use it myself, but a colleague swears by Acon DeVerberate. https://acondigital.com/products/deverberate/

[Tomás Mulcahy]

for the main software and then add the UVR-DeReverb-aufr33-jarredou_4band_v4_ms_fullband algorithm from the VR Architecture in the Process Method dropdown.

Thank you for the tip, I must try this. I've had good results with UVR-De-Echo-Normal, which I started using because it can deal with echo as well as reverb.

[James Perrett]

Thank you for the tip, I must try this. I've had good results with UVR-De-Echo-Normal, which I started using because it can deal with echo as well as reverb.

I suspect that the best algorithm is very much dependent on the material. When I fired up UVR to get the exact name of of the model that I use I saw that there was another algorithm on the list that I hadn't noticed before - developments are still fairly rapid in this area.

[Tomás Mulcahy]

Yes, I could not find the one you mentioned. I would recommend using Audiostrip for the stems with UVR for additional de- reverb/ echo. Audiostrip is very easy to use and sounds better than the Demucs 4 in UVR, plus it can give separate drums, and is pretty good at separating lead from backing vocals. It’s better than the Steinberg and Apple offerings too. I’ve not had much success separating piano from guitar with any of them though. It’s amazing tech, as with James’ example I mostly use it for tidying up old live tapes, where the mix was usually straight off the desk with nothing of the room- so you only get what was run through the PA to reinforce what was already coming off the stage.

[James Lehmann]

I use Izotope RX De-Reverb for this task.

As always, it depends what and how much you are trying to remove in order to assess the trade-offs.

Most of the time here I'm trying to reduce unwanted room tone on raw vocals & instruments, and thus use RX sparingly for incremental gains for which it works well. As you can imagine, I got quite adept at deploying it over lockdown!

I have no experience using RX for suppressing large amounts of artificial reverb on a full mix.

And obviously it's probably an expensive 'one-shot' solution unless you can justify buying in to the whole RX Suite of NR tools in a professional situation.

[James Perrett]

Yes, I could not find the one you mentioned.

I may have downloaded it from somewhere like Github.

I would recommend using Audiostrip for the stems with UVR for additional de- reverb/ echo. Audiostrip is very easy to use and sounds better than the Demucs 4 in UVR, plus it can give separate drums, and is pretty good at separating lead from backing vocals. It’s better than the Steinberg and Apple offerings too. I’ve not had much success separating piano from guitar with any of them though.

I've heard that the latest stem separation in Logic is based on a fairly recent algorithm though, as I don't use Apple, I haven't had a chance to try it. Piano is one thing that Spleeter does reasonably well - but you'll need to dig around a bit to find the software.

Are you keeping up with this document?

https://docs.google.com/document/d/17fj ... iuj54hzww3

It looks like there are loads of new models coming out...

[Tomás Mulcahy]

Oh, brilliant thank you, was not aware of that! @ James Lehmann, FYI UVR with that algo is vastly superior to RX de-reverb.

[Gutter Puppy]

Thank you all.

Looks like I might be out of luck . . .

Just to clarify/summarize -

I'm dealing with single wav files - each containing either a song recorded "live" on a 2-track/stereo tape/CD recorder, or the stereo mix of a song previously recorded on one of the various multi-track machines I once (but alas no longer) possessed.

Perhaps being overskeptical here, but how can/could the aforementioned software effectively separate/divide a multi-instrument/vocal/compression containing mix (eg. vocals, drums, bass, acoustic/electric guitars) into distinct, individual, sufficiently sonically and EQ differentiated (for reverb removal purposes) stems, which, post-removal, may be recombined without affecting the musical integrity/sound/tone of the original mix ?

[The Elf]

Try it. Be realistic and accept that it's a compromise. These processes are not as good as you might wish for, but probably better than you fear. When I've had to use them (e.g. Steinberg's Spectralayers) it's been because I've had no other choice - and the results have usually been good enough to scrape through. Just don't expect miracles.

[James Perrett]

Thank you all.

Looks like I might be out of luck . . .

Just to clarify/summarize -

I'm dealing with single wav files - each containing either a song recorded "live" on a 2-track/stereo tape/CD recorder, or the stereo mix of a song previously recorded on one of the various multi-track machines I once (but alas no longer) possessed.

Perhaps being overskeptical here, but how can/could the aforementioned software effectively separate/divide a multi-instrument/vocal/compression containing mix (eg. vocals, drums, bass, acoustic/electric guitars) into distinct, individual, sufficiently sonically and EQ differentiated (for reverb removal purposes) stems, which, post-removal, may be recombined without affecting the musical integrity/sound/tone of the original mix ?

If you haven't experienced this software then I can understand your scepticism. Indeed, the first time I tried it, it seemed like magic. But UVR is free and you'll lose nothing by giving it a try. The Purple Hearts recordings that I linked to came to me as stereo mixes and they ended up being spread over something like 10 tracks once I had done all the separations. Granted, some of the separated tracks didn't sound great on their own but they gave me access to modify the sounds and create something dramatically better than the mixes I had been given.

[Forum Admin]

...how can/could the aforementioned software effectively separate/divide a multi-instrument/vocal/compression containing mix (eg. vocals, drums, bass, acoustic/electric guitars) into distinct, individual, sufficiently sonically and EQ differentiated (for reverb removal purposes) stems, which, post-removal, may be recombined without affecting the musical integrity/sound/tone of the original mix ?

It's called Stem Separation, and relies mostly on AI. There are many tools out there that offer separation of sounds... have a read of this SOS article:

https://www.soundonsound.com/reviews/zplane-peel-stems

[MOF]

or the stereo mix of a song previously recorded on one of the various multi-track machines I once (but alas no longer) possessed.

Do you still have the multitrack tapes, you could get them transferred by companies that specialise in that?

[Forum Admin]

Do you still have the multitrack tapes, you could get them transferred by companies that specialise in that?

This Forum's very own James Perrett offers such a service, through his studio business.

[BigRedX]

I've just used the Stem Splitter function in Logic to separate out a rather poor second generation cassette copy of a live off the desk recording of a band I was in that was made almost 40 years ago. Two of the songs have elements in them that are worth revisiting, but it was such a long time ago I was struggling to identify all of the notes I was playing. Apart from a couple of instances where it had problems discerning the difference between overdriven guitar and overdriven bass, it has done well enough for me to be able to work out the tricky parts. As a bonus I've also got an isolated drum track that I can use as the basis for programming something more in keeping with my current band's rhythmic style.

[ManFromGlass]

A few months ago I, too, ran a cassette that was about 40 years old through Logic’s stem splitter. My expectations were low but the results were mind boggling.
I did a remix with better verb and panning etc and sent it off to a buddy who loved the walk down memory lane.

The only downside I found was on hearing old material it made me think I used to be more creative back then - back when I didn’t know so many of the “rules”!

[Tomás Mulcahy]

As with BigRedX's story above, I've used it to finish or develop or learn parts from old demos, often while retaining some of the original elements. For me, now is the best time because I know how to finish songs

[Mattsong]

I had good results on a 50 year old real to real tape. I tried it with the older Stem Splitter in Logic and it was pretty good, but a low synth part was unusable. The new Stem Splitter, that was updated for archival recordings, did a much better job and the synth bass was very good.
One interesting aspect was that the vocal stem was stereo for the verses, but for the chorus it became, lead vocal on the left channel and background vocals on the right channel. Has anyone else noticed this?

[jimjazzdad]

While I have enjoyed following this thread - it's always nice to meet old Fred Grift again - am I the only one who read the OP's request ("effective at reducing/removing reverb from an established/final mix") as being about some magic tool one might plug in on the 2-bus, rather than splitting-out tracks and reconstructing a mix? For example: taking away the cafeteria 'verb from an old tape of a live performance; removing the 'small room sound' from a piano recital to replace it with a lush Bricasti...etc...etc. I for one would be interested to know what people have tried.

[James Perrett]

am I the only one who read the OP's request ("effective at reducing/removing reverb from an established/final mix") as being about some magic tool one might plug in on the 2-bus, rather than splitting-out tracks and reconstructing a mix?

I read the request the same as you - but I felt that reverb reduction was more likely to be effective when applied to individual stems rather than the mix as a whole.

Of course, if we are talking about solo piano then there's no reason to separate into stems. Just use one of UVR's de-reverb algorithms - although, if the intention is to add a much larger reverb, sometimes the de-reverb step isn't needed.

[ManFromGlass]

I was leaning the same way. I discovered in my scenarios that verb was not a predominant part of the sound and adding new verb to the old verb worked. But I got lucky.