autotune and avatar

admin — Sun, 24 Jan 2010 02:32:20 +0000

Although autotune is now used as a digital effect, it was originally used to correct pitch in songs. With its use, singers can sing in perfect pitch, so long as they are not too far off. Indeed, autotune does not need to work in real time, and at a high level, it is no different from an instrument synthesizer, but with the instrument sampled in real time. (Perhaps such a hybrid approach could render even more realistic real acoustic instruments, and make almost anybody a “great” music player.) As the automated portion of the autotune’s capability improves, less and less of the singer’s input is needed, and one finds less and less need for the perfect singer, and more and more need for the perfect song and its performance intention — this is, after all, the essence of a creative work — not the much valued virtuosity with which it is performed (for its “difficulty”).

A similar thing has been taking place in motion picture production, with computer assisted graphics taking over for effects and stunts. Lately, the production process for the movie Avatar has pushed this process to a mini-plateau of some sort. Avatar, as you may recall, is produced by sampling the expressivity of the actors on a body grid, then re-rendering in a very different way. Much like the human-controlled machines in the movie, the actors are just giving input to a machine, which follows the director’s desires. Again, as the technology improves, less and less of the actor’s input is needed, and eventually, they, like the singers, will be unnecessary.

When it comes to the creative fields, as it does — I believe — in any field, the evolution of automation technology diminishes birth advantages, allows compartmentalization of skills, promotes specialization of skills, and therefore equalizes opportunities. The beneficiaries are people who engage in true creativity of the mind, both in the arts and in the engineering of the technology, while the losers are the human “performers”, save for the few truly great ones, who will be needed to go through the dehumanizing experience of being sampled as input for a machine.

So… study what a machine cannot do, or, study how to make a machine do that.

Detecting true perfect pitch

admin — Mon, 08 Jun 2009 19:58:11 +0000

This article (also this) proposes that there are two types of perfect pitch, “ability to perceptually encode” and “heightened tonal memory”. And these groups perform differently on a tonal matching test. I take the first to mean the ability to match any tone whatsoever precisely, while the second one to mean the ability to have long-term memory of certain heard tones.

It is interesting to consider what kinds of test actually measure perfect pitch. Usually there are two abilities under consideration, one is the ability to recognize heard tones by their names, the other to generate tones upon calling their names. The proposed article seems to say these two in themselves are rather symptoms of either APE or HTM or even something else as manifested in an association task. Indeed, the recognition task (hear a tone, call a name) is not strict enough to identify either APE or HTM. A piano player may have tactile or visual idenfication of heard tone with position on keyboard, and mediated by this association, know the name of the note — although this is usually not the case. Same goes for all the tests involving reproducing a note on an instrument or using vocal chord position, etc. These are cases of a “hidden” external reference. The mediating step is not seen. The generation task is more interesting, as it must involve at least tonal memory in the form of an internal reference. If it can be done accurately then it could be either APE or HTM but it would not be able to distinguish between the two.

The test proposed by the article solves some of these problems by requiring generation, and by using distraction after the short target tone is produced. The point is to move on from the target tone faster than consultation with hidden external references can take place. If recognition is not immediate, then one must first hold the note in short-term memory, then after the distraction, compare it to internal reference pitches from tonal memory. This is not accurate since short-term tonal memory itself is not stable, being influenced by distraction. So for some small number of tones (could be all of the chromatic scale), HTM could do well, depending on the person, but maybe performance is not even…, and HTM should never be able to match lesser-heard (e.g. non-standard) pitches well… However, if recognition is by APE, then any tone can be immediately recognized into an abstract form and as something distinct, and easily matched later in the abstract forms.

Under this regime, it would seem that most people who recognize and generate tones upon request probably just have varying degrees of HTM and have developed a quick lookup table as internal reference, which would seem to be malleable by training as with other kinds of memory (for people with good associative memory anyway). APE, however, probably cannot be learned — it’s a kind of idiot savant skill like people who know large number multiplications in one second — it just cannot be done with a lookup table.

P.S., here is a highly enlightening thought experiment by somebody trying to learn perfect pitch, and I must say it expresses almost perfectly my thoughts on the subject.

Some stuff » perfect pitch

autotune and avatar

Detecting true perfect pitch