Here is the third sound in my installment of FM synthesis experiments. Once again I’m using artificial selection to simulate a genetic algorithm. In this case I selected a rhythmic sound out of a large collection of randomized sounds then repeated the two bar loop four times to create an eight bar pattern based on the sample. To me it sounds like some sort of high tech laboratory equipment being operated.
One quick note. I will be out of the country from Friday, November 7 through Wednesday, November 12, 2008. On Friday, November 7 I’ll be posting my 100th consecutive sound, and want to continue the trend while I’m out of the country. So, I’ll be scheduling a series of entries based on these FM synthesis experiments to post automatically while I’m away.
In my last post I explained how I rein in random processing behaviors to get the results I’m after. A good processor for randomizing audio is Ableton Live’s Beat Repeat. Beat Repeat effortlessly duplicates the once tedious process of repeating small chunks of a sample to get stuttering effects, but also has parameters to randomize the repetitions in a variety of ways.
For the Rhodes solo in “Six Weeks” I wanted to scramble my performance in some way to match the “broken” drum programming. Beat Repeat was the ticket, but I couldn’t get a complete take that fit well with the rest of the piece. If you look at the image you can see that the solo is made up of fifteen separate regions of audio. These are all abstracted from specific renders of the performance through Beat Repeat. After rendering the audio several times I selected specific phrases and organized them in a way that enhanced the dynamics of the piece, creating a hybrid. Listen to the solo by itself and then to hear it in context play the full track at 2:54.
After playing Precambrian Resonance for a few people and explaining how the arpeggiator was creating a randomness to the output I was asked how that randomness made it sound different from previous playback. This was easy for me to imagine since I had heard it rendered several different ways, but difficult to explain. Therefore I have re-rendered the piece to illustrate how it changes.
This brings up an issue that I have encountered on several occasions. When audio processing creates some sort of randomness in a mix, how can you get exactly what you want? What if after you export the audio there’s some chunk of randomized audio that just doesn’t quite work?
My solution is to render the track that has the random processing on it several times. For Precambrian Resonance 0.2 I rendered the processing eleven times. After that I’ll listen and compare the renders, or if I hear one that I like during the rendering, I’ll just choose it. Ableton Live makes this easy with the “Freeze Track” option that essentially renders the track while allowing you continue making adjustments.
Sometimes it is not that easy. I have encountered situations where version after version of the randomized processing doesn’t quite fit. At this stage what I do is carefully listen to the audio for phrases that have something interesting going on. The next step is to sequence the selected phrases into a complete track, effectively herding the random behaviors into what I’m after. I suppose that this is similar to using genetic algorithms to hybridize the audio in a semi-manual way.
Here is the third sound in my installment of FM synthesis experiments. Once again I’m using artificial selection to simulate a genetic algorithm. In this case I selected a rhythmic sound out of a large collection of randomized sounds then repeated the two bar loop four times to create an eight bar pattern based on the sample. To me it sounds like some sort of high tech laboratory equipment being operated.
After playing