Can any of you fellas tell which is better?

 

Just did a quick test of Logic 11 against UVR5, only vocals and instrumental. UVR MDX NET sounds way better, also the sum of Logics stems doesn't null to the original, I'm not impressed at all. Haven't tested RX 11 yet.

 

UVR5 best at the moment

 

+1 for UVR5, MDX23.
Definitely not Acoustica, it's ass.
No opinion on RX/Logic.
People claim FL's separation is surprisingly good, won't check myself.
Ripple (iOS only, not available in the EU) scores highly on leaderboards, but can't verify.
If a truly simple solution with a GUI is needed, SpectraLayers is comparable in quality to free online AI services - usable but not exactly mind blowing.

 

UVR5 with HTDEMUCS_6S is miles ahead of RX 11 Music Rebalance, and Logic Pro 11 most likely just implements what UVR5 had for long time...

- that way you get 6 separate stems (drums, bass, guitars, pianos, others, vocals)

 

Acon Digital does a great job at vox and drums.

 

You mean Acon Digital Remix? Have you heard ANY of the above mentioned algorithms? They are all miles ahead, including Logic.

 

I haven't heard of those. Are they open source?

 

UVR5 is: https://ultimatevocalremover.com/

 

Is this a true open source forever, or will they start selling for $99 after an official release? That's happened so many times before.

 

it's just a GUI frontend for various AI models like MDX, htdemucs and such, in its advanced settings you can download tons more to try

 

i've been getting great results with RipX, its much more granular as well, rather than being limited to 4 tracks

 

Yes, they can't take it away anymore. Same for the under-the-hood models. Anyway, you quit for the seventh time this year, what's it to you!

 

has anyone compared these with stemroller? SR is free and open source

 

Haven't tried it yet, but it is based on Demucs, so it should be comparable to one of the Demucs models in UVR.

 

Neither. Go with UVR5.

 

I downloaded and will certainly try. What, in your opinion, makes UVR5 superior to the countless others? Less artifacts? Cleaner?

 

Try UVR5 with MDX23 model and hear for yourself. It's just vocal and instrumental and takes a lot of processing power, but it is very clean with very low artifacts. Also delay and reverb are retained better than most other models do. As tzzsmk stated, htdemucs_s6 gives you 6 stems, but vocals are not as clean as MDX.

 

Pretty much all of the above. But don't listen to me. Do a test for yourself.

 

UVR5 with it's MDX and RoFo models, also some of the Demucs models. all very solid seperation models.
if you want clean seperations and all that, then you'll ofc have to put in some work! like, doing an extra seperating of seperations.
or even further, if you want specific seperations, you'll have to train your own models, which isn't difficult but sort of alot of work!

i say 2-3 more years and you'll be able to seperate alot from eachother. it's just the immense work of model training!