António Florença

Long forms are only meant to express here musical form across several bars and whole movements, i.e. composition as it expresses itself in long spans of time. Once again it seems that I'm hitting the wall hard on how well I can express my thoughts publicly, but here goes:

What techniques are there in Computer-Assisted Composition (CAC) in order to help us sculpt/interpret the long form of musical pieces? I've struggled with this over the years, I know and employ several short-form techniques in my own compositional boutique but I've never been able to conceive of any long-form techniques.

Any hints and insights or free-floating thoughts on this issue would be much appreciated.

Hi,
Long forms are only meant to express here musical form across several bars and whole movements, i.e. composition as it expresses itself 

long-forms-in-cac

the math for analysis and generation of data is somewhat the same than for short form, no? there is time and there is pitch and there is change, similarity and repetition...

what can be helpful in my experience is that you include processes like averaging, finding the most common stuff and ignore the special cases, the longer the thing you want to analyse is.

generation? i dont know. for me, changing the key and BPM after 7 minutes is not something i as a composer would like to leave to a machine. it too much depends on things like sound and listening context. sure, today you can pretend all that by using AI, but nah.


the math for analysis and generation of data is somewhat the same than for short form, no? there is time and there is pitch and there is change, similarity and repetition...

what can be helpful in my experience is that you include processes like averaging, finding the most common stuff and ignore the special cases, the longer the thing you want to analyse is.

generation? i dont know. for me, changing the key and BPM after 7 minutes is not something i as a composer would like to leave to a machine. it too much depends on things like sound and listening context. sure, today you can pretend all that by using AI, but nah.


Now, AI indeed won't solve your question, buy having the computer in the loop could be fun... but it will require as anything a human in the loop too :)

For instance, analytically, you can look at similarity graphs to observe formal returns. You could feed proportions/states/behaviours/sections into a Markov chain, and when you are back somewhere you can distort/mutate instead of copying.

Obviously, the Bach suite is the place to go to implement such things. I reckon you know it but if not, Cage and Bach are your friends. The tutorials are fantastic too. 

Obviously, the Bach suite is the place to go to implement such things. I reckon you know it but if not, Cage and Bach are your friends. The tutorials are fantastic too. https://www.bachproject.net/ 

Thank you both for the great pointers, and I completely agree when it comes to Artificial Intelligence and Machine Learning in CAC.

Thank you both for the great pointers, and I completely agree when it comes to Artificial Intelligence and Machine Learning in CAC.

All the best, 
António


Long Forms in CAC