[PD] Turning non-audio data feeds into audio

[Mathieu Bouchard]

On Sun, 11 Oct 2009, Jerome Covington wrote:

> Is anyone interested in sharing their process for turning real-time, 
> non-audio data feeds, into music? See a great example of one possible 
> direction, here.

Coïncidentally, I wrote some thoughts about it in the Pd chatroom a few 
hours before your email, because of a similar topic there:

«

musical meaningfulness comes from meaningfulness of the data beforehand... 
basically, if you put garbage in, you get garbage out.

the exception to that is that a programme is a kind of data in itself, so 
the programme can be considered a kind of meaningful input... and if the 
programme imposes itself as the source of the meaning and successfully 
downplays the incoming garbage, it can make the output meaningful;

but unless one is very skilled at understanding the information theory 
standpoint of music, using random values gives you just more meaningless 
music like what you are talking about... sort of like picking a random 
book from the library of babel.

»

> http://vimeo.com/5415629

now this is what I add to my above thoughts, this time in relationship to 
the video: without necessarily explicitly thinking about information 
theory, one can get to interesting results intuitively... one essentially 
has to focus on getting beautiful results for likely inputs instead of 
being content with whatever fits with the description of a certain art 
concept. Any former stock-market music I had listened to sounded like 
crap. What Patrick did was to make his programme insert so much beauty and 
coherence in the market's noise, that it made it sound meaningful... 
actually, it's more like this: the programme can only output music that 
sounds reasonably good no matter the input, and the meaningless input 
selects one of the possible nice-sounding outputs. Overall, the music is 
more shaped by Patrick's æsthetic decisions than by the stock market, and 
it's perfect like that.

so, Jérôme, I would mostly just suggest that you make patches so that the 
results sound fairly good no matter the input you give them, and 
optionally, if you can make the input also recognisable in the output, 
it's a bonus feature that can feel very rewarding, but it depends on the 
context... for feeding stockmarket data it may not matter as much, but for 
live interactive data from performers or visitors, they have to recognise 
their own impact on the music, else the point is going to be lost on them, 
really. but even for stockmarket data, it's better if you can recognise 
the stock price in the music, because if you can't, you could have taken 
that data from anywhere else and it wouldn't matter, so why would you call 
it stockmarket music then?...

so maybe you wanted people to explain their actual processes, but I hope 
that you will also enjoy this reflexion on the question of what might make 
processes be good or not.

  _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard, Montréal, Québec. téléphone: +1.514.383.3801