PG Music Home Forums General Discussion Off-Topic Are there any things you do to "humanize" your generated Tracks ?

Ah.

Joe, I think you are confusing some terms here.

Think about this:

MIDI synthesizers, whether hardware or software, depend on Samples made by recording or synthesizing each possible note in pcm digital audio. When sampling an instrument for such, the job is painstaking if done well, and just recording each note can take more than just time. For example, I once was consulted to design and build a programmable electric "finger" for the purpose of "playing" a single piano key at a time, but with exact velocities and pressures. This was one way of obtaining multiple samples for the Velocity Layers that the engineers wanted to obtain from the great sounding grand piano, mic'd well and in a great sounding acoustic environment of the studio. When the amount of layers desirable were more than three (Soft, Medium and Hard) they soon found out that use of a human being to hit each key was problematic. Part of the issue also turned out to be due to the fact that the amount of pressure, or velocity, needed to get the same layer was very different from the lowest key on the piano to the highest, and by building the programmable finger we also learned that the differences needed across the keyboard weren't even a linear progression. I also learned that this whole thin can differ from piano to piano, for example, the famous Steinway "Accelerated Action" pianos required a completely different programming template for our laboratory grade robot finger than was required to sample a Baldwin or another brand and get equality of sound.

Now, that's just the example of a Piano, in which the Attack of the notes are really governed by the Hammers.

In the case of many other instruments, horns, guitars, woodwinds, strings, etc. the ADSR (Attack, Decay, Sustain, Release) of the sample set governs what you hear.

If a horn were sampled with the horn player tonguing each note, then there will be a "Ta" kind of Attack at the beginning of each sample. Great for emulating a Bugle Call, would likely be good for the start of the William Tell Overture, but since there are no samples there for emulating a slurred passage, in which maybe only the first note in the phrase is tongued, the following notes being articulated by valves and lips, attempts to throw the same MIDI sample at it would result in every note sounding as if it was tongued, regardless of what can be done with CCs and such. It thus could not sound realistic.

I will put this into terms for the guitarist now.

A MIDI sample set of a guitar would have the same problems in ADSR, if the samples were made of all picked notes, then there would be an Attack at the beginning of every note fired by the MIDI file.

But we often don't pick every note we play, right. Hammer-ons, Pulloffs.

The MIDI sample with a pickmark at the beginning of every note cannot sound like that because every note will have an Attack that is greatly different from the desired sound.

Over the years there have been various attempts to make things better by using workarounds inside the aging MIDI Standard. Some have touted custom uses of the CC's such as using the Modwheel to allow a keyboard player to access different samples on the fly (with a little practice in playing the keys while working the Wheel) in efforts to emulate string bowings on the fly, or guitar playing with pick attacks, hammerons and pulloffs. Garritan's offerings are one example. BUT -- use of nonstandard stuff for CC's also means that one cannot simply load up any old MIDI file found and get it to playback what the original creator of the file, likely using a GM bank, heard.

Okay, that's just a quick explanation of the Attack problems inherent in firing samples via MIDI.

There is still the "DSR" to be considered, the Decay, Sustain and Release aspects of each sound.

These too, can be different depending upon other issues, the Attack for one. When you pluck your guitar string lightly, the DS and R are sounding differently than when you PLUCK that string with vigor.

How about the different ways that we can pluck those guitar strings?

*Was the pluck done with a plectrum? If so, what kind of plectrum, what particular qualities does the material, thickness, stiffness, flexibiltiy of said plectrum bring to the sound, not only of the Attack, but the entire note? Bear in mind that for a MIDI synth, you could sample the Attack in layers, perhaps the DS&R in layers as well, but again here we get into how much a particular MIDI synth solution costs vs that pesky Law of Diminishing Returns.

*You are into Flamenco guitar, so we are likely not even talking a plectrum in that case. While you might hear and see plectrum in use today in certain Flamenco situations such as soloing, you know darn well that the true Flamenco player uses picking hand and fingernails. The four finger strum sound, when executed well, is Flamenco signature of a sort. Got any MIDI samples of that? How would that be able to be implemented in a MIDI file? Here we get into specialty samples, which are actually available, and the clever MIDIOT might take advantage of such by initiating Patch Changes in order to shift from, say, a bit of PIMA over to the frail and then Patch Change again to get back. Again, such is more expensive to have in the average home recordist's arsenal. (Studios such as Dreamworks, however, can and do afford to do things like sample the London Symphony all kinds of ways, but their samples are proprietary and belong only to them to use. But they can justify the expenditure over time, because they also do not have to hire 80 to 120 orchestral musicians to do every underscore anymore. They do it with MIDI. And most folks and most musicians cannot tell the difference.)

The MIDI Standard as we have it today can only accomplish so much.

However, I would caution anyone who is thinking they could come up with something better, because they would only end up adding many more commands, samples, Continuous Controllers, perhaps a few other things in which a terminology would be required -- but the MIDI Standard we already have confuses most to the point that they lose interest quickly enough due to complexity. So they declare that MIDI sucks, demand real audio loops and more Realtracks to make their home productions, and then its off to the Movie Theater where they are confident that the MIDI they are hearing there is a real live recorded orchestra. Or really great Concert Grand Piano in an equally great hall. Sometimes it is, quite often it isn't.


--Mac