AET feedback?

I'm very excited about this new feature but i don't see much talk about it, anybody tried it?

 

yes i tried it, but without any useful result. think i´m doing something wrong, although
i used the (baaaad) manual.

 

Bab manual huh? I bet that doesn't help... but from the video demonstration it seems fairly easy to apply AET, did you try that?

I'm surprise not to see more interest in this...

Cheers.

 

During the Beta testing phase I spent many hours trying to get spectacular results from the AET and some experiments yielded interesting results. But it's not really designed for morphing sounds into each other rather than smoothening the crossfades when using multi sampled Instruments. So the AET takes a footprint of the different spectral structures of different samples and tries to morph them into each other. It works quite well when used in multi velocity layer mode for treating multi velocity drum and percussion sounds.

 

Sounds good. Smoothing the crossfade of multi layered instrument is exactly why I'm interested in AET and K4. I mostly want to use it for orchestral instruments like Brass/woodwinds and solo instruments like violins etc.. You think it will work just as well for these instruments?.. I didn't realize this could be useful on percussion. Interesting...

Cheers.

 

Yes, morphing different velocity layers of brass and string Instruments with e.g. the Modwheel should yield good results and I think that's the main reason NI invented this feature. I haven't taken the time to modify my orchestral Libraries to try that out and actually I'm quite happy with my VSL lib the way it works now so I'll rather go on abusing AET for my needs.

 

Great!

Thanks.

 

I've tried out the new AET and offer the following comments.

First of all, contrary to what is implied in the blurb it doesn't morph from one sample to another. What it does is to superimpose ("morph") sound characteristics from one sample onto another in a gradual manner. So if like me you were looking for a means of getting a crescendo on a french horn gradually changing from an mf sample to an ff sample, forget it, at least at this stage in the development. The sound resulting from 100% morphing of the ff sample onto the mf sample is not useful: it is distorted and is nothing like the ff sample. If you want to get the ff sound you have to start a new note, which defeats the object of the exercise.

I hasten to add that I use a breath controller to modulate volume, with a fixed velocity, but if you play your orchestral notes with multilayer samples on a keyboard and modulate volume with velocity, you could well find it valuable, though of course you won't get any useful crescendos etc within notes.

And if you want weird noises, there's plenty of scope for that.

One or 2 details.

1. You are prevented from creating 2 morph layers containing the same zones. This is unnecessary since Morph Layers don't do anything until incorporated into a morph map and it would be useful to have a collection of layers with different settings for experimentation. (You can get round this by duplicating the zones, but this shouldn't be necessary).

2. You can't have zones from more than one group in a layer, even when zones from several groups are selected during creation. (Bugfix please!)

I'm hoping the technology is still under development. It would be terrifically useful to be able to truly morph from one sample to another, so that at 100% morphing you were playing purely the 2nd sample. I'm avidly watching this space.

Stephen

 

Thanks for the detailed response stephen.

Yes, I'm in the same position as you. I want to be able to cres using different sampler layer like p, mp, mf, f would be ideal. I'm a little disappointed cause i get both + and - feedback on this. But jugging from the sax example in the video presentation, the result seem quit good. I guess it doesn't always work well if it didn't work well on your solo horn...

I hear the benefit of doing things this way is that it trigger 1 group/2 voice as oppose to 4 group/8 voice in the usual mod wheel velocity crossfade instruments.

Feel free to try on other solo instruments like Woodwins and brass and report back if you want?

BTW, how is K4(AET) cpu wise compare to K3? If you don't mind me taking more of your time.

Thanks.

 

I have the same opinion as steven22.

AET is a dissappointment. Either the technology isn't capable of morphing between sounds, and this is all a big marketing joke, or the technology is working, but the implementation (and especially the documentation) is far beyond acceptable.

The morphing-demos as well as the choir-patches show that there is some potential, but unless NI throw out a real tutorial how to blend sounds (I guess there has to be some audio-editing before importing the samples in K4), it is a non-usable technology.

If someone else made different experiences, I would be glad to be proven wrong, since I bought Komplete 6 mainly just because of this feature.

PS: I think the CPU-load is approximately as high as the convolution effect.

 

Thanks Chris.


I'm very happy NI decided to go that route because i think this is the one important missing feature. Of course this is an advance sampling feature so i guess it's understandable the result aren't exactly there yet, but i hope NI will keep going in that direction.

Money is tight right now so i may have to hold off on buying K4 since AET is the one feature I'm interested in. I'd be curious to hear what NI has to say about this at this point.

 

many thanks to Stephen and others for the detailed insights.

 

just bumping the thread here..!
i am not a very advanced user .. video instructionals would help us get on our feet faster and also help us understand what was the intent of the programmers /designers to introduce this feature and how they expect us to use it.
And as this is not just about dropping LFOs and ENV.s on parameters .. we would really appreciate a little head start .
NI has been actively posting some great videos of other products like mashine and kore on youtube..
would love to see one on AET as well.

 

I just got Kontakt 4 a day ago and what I found so far is with the velocity map it will sort of morph between samples where in multi velocity layered instruments smoothing out the differences between instrument layers. This works pretty well.
When I tried to use a continuous controller on a sax sample I got it to work somewhat getting crescendos sometimes and decrescendos on other times. The devil seems to be in the details here I'm going to need some work with this. I'll get back to you if no one finds the answer first. I have not gotten beyond velocity. Try it on a bass sample or some instrument that does not have a lot of layers. I'm curious.
BTW to get it to work I opened up the reference manual and knotakt at the same time and followed it step by step

 

I would like to elaborate on what I said and perhaps clarify my thoughts a bit.

What the new technology does is to superimpose some of the spectral characteristics of one sample onto another in a rather clever mathematical way. The original sample is always there, but the second sample is "morphed" on to it to a controllable degree. The program uses this in 2 distinct ways:

1. if the instrument has multiple velocity layers, it can do a high-tech "crossfade" between the layers. So a sample in the "p" layer will have spectral elements of the "mf" layer superimposed on it, and as (with repeated notes) the velocity approaches the boundary between the layers, the amount superimposed increases to a maximum. When the boundary is crossed the reverse happens, and a variable amount of the "p" layer is superimposed on the "mf" samples.

2. A set of samples may have spectral elements of one or more other sets of samples superimposed on it, the amount superimposed being controlled by a modulation source, usually a cc. So you can have a set of "mf" samples and superimpose "f" samples on it, to a degree controlled eg by the mod wheel.

This is all entirely adequately described in the manual, and though it's not entirely clear what an attribute called "smoothing" does, I suspect this is a mathemetical rather than a musical parameter and its effect will only be determined by experiment.

It seems to me there are 2 issues: the immediate one being which sorts of samples will "morph" together in a useful way. This can't be learnt in tutorials but only by individual experiment and I for one would be most interested in hearing about peoples' results (as I already suggested, the creation procedure could be modified to make this easier). My own example of an mf/ff french horn suggests that samples with greatly differing spectral characteristics probably won't work well.

The second issue is possible development of the technology. I'm going on about this a bit because it's the major new feature of Kontakt and I think extremely exciting. In its present state it doesn't actually morph between samples in the true video sense of the word (seamlessly sliding from one sample to another) but it seems to me this shouldn't be too difficult to achieve.

Stephen

 

What smoothing usually means in Kontakt is that you can go beyond the midi specification of 128 step to eliminate the zipper effect when using a pitch wheel for example. So if you set smoothing to 250 you get 250(+128) = 378 step. You basically get better quality but it will add some latency(lag) by the same amount. 250 millisecond in this case. I hear 250 is a good value for smoothing mod wheel instrument BTW.

I assume smoothing will have similar effect here and the higher you set it up the better quality you'll have etc..

Brass is THE one orchestral instrument category that could benefit from AET the most so I'm a little disappointed. Maybe you didn't add smoothing? If so, I'd really like to know if it made a difference?


Thanks.

 

Might have a different meaning here. To quote the manual:

I did and it didn't.

 

Finally there is some communication about this feature. It's somehow strange that there is no comment from NI although every other thread is almost immediately answered ("HELP, my library isn't installed"...).

I have an (theoretical) idea: maybe the spectral characteristics are not too different but too similar. If you superimpose a characteristic on another, some existing frequencies (especially the fundamental tone) are doubled, since they exist in both samples. This might lead to the distorting behaviour I recognized in my "researches": when superimposing a clarinet "ff" on a clarinet "pp", the frequencies which get amplified are the fundamental note and top end breath noises, but not the uneven harmonics like it should be.

The solution could be some sample-preprocessing: if you could somehow substract the pp sample from the ff sample (maybe by some intelligent FFT filtering), you would not double some frequencies when superimposing.

I can't test my theory this week, but maybe someone else gets inspired by this thoughts. Together we can make it!

 

it is important that when morphing different samples that no frequency gets deleted or that aliasing effects occur. I don't have Kontakt 4 yet to test AET, but I have a special breed of Kontakt libraries that can morph between layers very nicely. It is from a company called Samplemodeling, and their technology is named 'harmonic alignment'.

You can not only morph between layers, you can also morph into frullato, growl, flutter sounds, and use different attack and bend characteristics. I do not know how exactly these morphs are achieved: are they real sample morphs or only modulations of samples?

But judgingf from the mp3 demos of AET it looks like both technologies (AET and Samplemodeling) are potent innovations to sound sampling.

 

You are talking about crossfading, not morphing. In fact, the mentioned problem is the main reason people are thinking about morphing-technology. I also own The Trumpet from Samplemodeling, and it seems to be a far superior technology to AET, although it's very cpu-hungry.