MeldaProduction

jmg8 · Post by **jmg8** » Wed Mar 01, 2017 3:16 pm

I also agree. Latency is not a problem. Really hope the feature can make it into the crossover. Especially if it could have a variable crossover, so we could even set multiple bands. Although I'd probably mostly use just 2.

Alphacodex · Post by **Alphacodex** » Fri Mar 03, 2017 7:10 pm

At first sight, I can really only see cool applications with an as artifact free as possible separation of tone and noise.

However there should be some tweaking possibilitys. iZotope has an "tonal/noisy balance" slider tweaking how much of the signal is either recognized as tonal or as noise.
At some other company I saw they had 3 gain sliders:
- tonal
- noise
- error
But I dont know what they do. Probably they have fixed crossovers and the "error" band is everything that could not clearly be matched as tonal or as noise...

At second glance multiple crossovers in melda-plugins could do both: 3 bands with variable crossovers. First band "tonal", the middle band would be the error / unsure, third band noise. If you want complete separation remove the middle band. Lowering that middle band in volume however could result in some interesting results, depending how broad it would be...

My phantasies go wild....

Alphacodex · Post by **Alphacodex** » Wed Mar 08, 2017 7:34 am

I found a scientific article describing an algorithm to do TSR (Transient Sinusoid Residual) splitting for audio.

https://www.hsu-hh.de/download-1.5.1.ph ... pRnLcQkzW4

There is also another method mentioned in the article called SMS (Spectral Modeling Synthesis) that only splits into tonal and noise and seems to work very well.With the search-term "Spectral Modeling Synthesis" you can find a bunch of scientific papers describing different approaches to this method and even some special applications (like vibrato exctraction with SMS)

I hope that is of any help Vojtech?

Cheers,
Codex

Alphacodex · Post by **Alphacodex** » Wed Mar 08, 2017 7:39 am

And especially for voice separation I found this:
https://hal.archives-ouvertes.fr/hal-01 ... 6-0722.pdf

MeldaProduction · Post by **MeldaProduction** » Sat Mar 11, 2017 10:44 am

Ok, so I got a little creative and we have 2 new crossovers - Spectrum that splits the bands depending on frequency levels, so highest band contains most prominent frequencies etc., and Tonal/Transient which apparently works very similarly to that thing from Eventide (which is not exactly a surprise, since it is a pretty obvious concept). Well, a secret preview beta for V11 should be available today/tomorrow,so...

mljung · Post by **mljung** » Sat Mar 11, 2017 11:15 am

Very interesting news - looking forward...

::
Mads

vectorwarrior · Post by **vectorwarrior** » Sat Mar 11, 2017 12:22 pm

Very exciting! Can't wait to try it out!

Curious to know if the spectrum mode accounts for weaker power of higher frequencies. Eg is there a slope control.

MeldaProduction · Post by **MeldaProduction** » Sat Mar 11, 2017 3:18 pm

And yes, slope is there, but it's called Tone

.

XComposer · Post by **XComposer** » Sat Mar 11, 2017 4:40 pm

Interesting! Looking forward to trying it (for the kind of use that I have in mind, I hope that the separation works not only with transients, which – in my case – does not matter very much to me, but also on any "noisy" part of any sound, also when the sound is stable, for instance on the air noise of a wind instrument, the friction noises of the bow on a string instrument, and so on).

vectorwarrior · Post by **vectorwarrior** » Sat Mar 11, 2017 6:59 pm

Yeah that's my thoughts too, we can already do transient detection so I'm hoping these two new crossovers offer something different. At the very least I think the spectrum crossover sounds really promising. Can't wait to test it.

Alphacodex · Post by **Alphacodex** » Sat Mar 11, 2017 8:10 pm

Yeeeeeehaaaaa!! I'm really looking forward testing the new crossovers? Where the articeles of any help?
Any chance for a third tonal / noise only crossover?

pumafred · Post by **pumafred** » Sat Mar 11, 2017 9:03 pm

MeldaProduction wrote:that thing from Eventide

MeldaProduction · Post by **MeldaProduction** » Mon Mar 13, 2017 5:05 pm

Hehe yeah I forgot the name

, my memory isn't what it used to be, and it was always bad

XComposer · Post by **XComposer** » Mon Sep 04, 2017 8:38 pm

Among the products already mentioned in this thread, did anybody of you test which one between iZotope "RX deconstruct" (a function within "RX Advanced") and the newer Sonible "Entropy: EQ+" (now included also in their "frei: raum") gives the best tonal / noise separation? I'm very curious (and I don't own any of them).

jmg8 · Post by **jmg8** » Mon Sep 04, 2017 9:27 pm

XComposer wrote:Among the products already mentioned in this thread, did anybody of you test which one between iZotope "RX deconstruct" (a function within "RX Advanced") and the newer Sonible "Entropy: EQ+" (now included also in their "frei: raum") gives the best tonal / noise separation? I'm very curious (and I don't own any of them).

Well iZotope sounds best, but that's not a fair comparison really as it's offline only. Plus there is no post processing, only separation. The other 3 are real time, in my opinion they all have similar results. But Melda is best for 2 reasons:

1. Eventide and Sonible have good separation for most material, but sometimes not. In this case there is nothing really you can do, as the separation is closed. In Melda's crossover there are 2 algorithms to pick from (tonal/noise) (tonal/transient) and then lots of control over parameters that determine how the separation is processed. Plus your not limited to just 2 bands, you can spread the separation over 6 bands. This makes it more useable on a wider variety of material.

2. What are you going to do after separation? Sonible let's you EQ the 2 separate parts, it sounds great but that's all it does. Eventide a bit more, it has 6 FX like EQ, delay, reverb etc. However Melda has 100 FX! Then when used inside MXXX, with it's routing capabilities, well, there is no comparison!

MeldaProduction

KVR Audio

Tonal vs Noise