Saturday, September 19, 2009

Sound Quality in Hybrid Production of MIDI and Recorded Audio

A publication level audio file containing recorded audio of real instruments and real players as well as MIDI sounds (or, more precisely, sample library sounds administered by whatever protocol) triggered by a sequencer is called a hybrid production. For the production of "orchestral" audio, given limited budgets, this is almost unavoidable under real market conditions (which means that, of course, if budget and time is no consideration, and/or if you or your friends deserve all kinds of instruments, it is possible to avoid MIDI altogether). Thus, in most cases, the question is not whether or not to go for a hybrid production, but rather how to make a hybrid production sound as good as possible. In a way this is what this blog is all about. Therefore, now I will fire off a couple of rules of thumb about how to avoid obvious inferiority of a hybrid production, compared to recording of great musicians in a great studio with a great engineer. This is only meant as a starting point rather than an exhaustive list, of course.

If the top/main voice is produced by a real instrument played by a skilled and sensitive instrumentalist, it is, under suitable circumstances, hardly recognisable whether or not the accompanying voices of the same instrument or instrument group (e.g. brass) are samples. I assume this is so because the static and stereotype character of library samples is veiled by the variance of the real lead voice. Some of its individuality seems to be psycho-acoustically attributed to the sample non-individuals. The more rhythmically homogeneous the section voices are, the more the samples melt into the musicality of the real instrument voice.

Thus, rule 1 is: if you can get a good oboe player, let him play melodies, leads, upper voices; this will help to hide a sampled clarinet and fagot.

Plus rule 2, from the same argument, states that, if you want to hide sampled brass, woodwind, or strings, you should let them play the same rhythm as your real player of the same section (or another; no limits!). As often as possible.

Instruments where the player influences the sound after initiating a note are much more difficult in MIDI than "percussion" instruments, i. e. instruments where, once a note is hit, nothing can be done anymore. This does not only apply to most percussion instruments, unpitched and pitched, but also to stringed instruments like harp, and even to pizzicato technique of violin, viola, cello, and double bass.

So this is rule 3: Try to rely on "percussion" instruments in the above sense as much as possible, if you are bound to use samples. Avoid instruments where the sound develops. To my ears, sampled singers sound especially awful, but saxophone and solo trumpet samples can also be really ridiculous (even if, taken as such, they sound great)

There is one important exception to rule 3: instruments a) which sound everybody is very much acquainted to, and b) where playing more than one note at a time produces interferences, or which lets a whole "instrument" resonate. Prime example: the piano. Play a huge pedalled chord on the (grand) piano, and listen... what you hear crucially differs from triggering, with correct velocities, these very notes in an expensive grand piano library. (There has been a good posting on this topic by Kenneth D. Froelich on his The Electric Semiquaver blog). Another example: a jazz ride cymbal. Here, it is not several notes at one time, but hitting the cymbal while it is still swinging, which makes it resonate in a way that cannot be matched by a one shot sample.

This leads us to rule 4: The shorter the sample, the better. Take the marimba as a prime example. The notes decay very quickly, and hence a good marimba library does a more than decent job. Exception: single notes; there is no problem in using a one minute gong sample because the whole rich resonance is part of this one sample.

Okay, one more, rule 5: don't be pedantic - if you do have a cello and do not have a violin, write low enough to make the cello play the whole strings section. This is what I've been doing a lot, being a cello guy. To me, that funny "cello section" still sounds better than violin samples.

Okay, one more, rule 6: don't be dogmatic - if you do have a trombone and do not have a trumpet, record the trumpet voice at half tempo one octave lower with the trombone, sample it, and playback at double pitch. There you are, sounds like a trumpet with a strange, fast vibrato, but way better than a prominent trumpet sample. That's what I did with Philipp Haagen for Schützenfest from the stage music for Die kleine Hexe.
blog comments powered by Disqus