Buf slicing comparison / learning patch

I’ve been looking at slicing buffers in the last couple of days and I thought it’d be cool to be able to see the affect of changing the many parameters of the slicer objects so I made a patch to compare:

fluid.bufonsetslice / fluid.buftransientslice / fluid.bufnoveltyslice

The patch lets you generate some weird techno, difficult fading section, load or record your own sound.

Once you have something in the buffer (1.)
(2.) choose which object you’d like to analyse the buffer with and after analysis the generated slices will be displayed over the top of the waveform view. (3. - The view can be zoomed and slid by clicking on it).

NOTES:

fluid.bufonsetslice deals pretty well with the Generic Experimental Techno source as I tuned the settings for it.

The difficult source (which has soft attacks and is intended to be a vocal analogue) is indeed difficult to slice and in the end found myself getting frustrated trying to understand the effects and interactions of the parameters. fluid.bufnoveltyslice does an ok job on it though it needs more parameter tuning.

Found something which feels odd about fluid.buftransientslice - if you click to do the analysis and then click again the markers will reposition which makes me think that unless i’ve patched it weirdly it’s giving different answers each time - Is this the expected behaviour? It doesn’t happen with fluid.bufonsetslice or fluid.bufnoveltyslice.

I’m pretty much tuning parameters in the dark at the moment so any insights would be appreciated!

fluid slicer tester.maxpat.zip (32.9 KB)

3 Likes

This is super handy as a visualizer.

I’ll play with this a bunch to see what works the best for me in a more “generic” use case where I slap it on everything.

I want to say there was another thread on here that had something similar (that @jamesbradbury had built?), though I don’t know where.

I haven’t looked at the patch in detail so you probably know this and have dealt with it already, but are you accounting for the fact that it spits out a start and end frame which are not onsets, but rather the delineated windows of analysis? Depending on what you’re expect that could have any zl group-type collection points spill over in an unintended way.

Thanks Rodrigo - I’ll need to check this, can’t remember if i did or not thb…

Here’s the thread I was thinking about. Some handy bits of code here that slice up what’s needed and works out a jsui visualizer.

This is great @leafcutterjohn. I had it on my to do list to make an abstraction that could help me with visualising slicing but now I don’t have to do it. It’s also way faster than jsui which is just balls when it comes to doing larger files. I like that this is zoomable too!

Thanks.

Actually looking at the patch now, and that’s not the case. I did a - 2 after the sampstoms~ to remove the last two points as they are not transients, and uzi ignores the first entry anyways, removing the starting boundary as well.

This is the relevant bit of the patch from the other thread that does the maths (though the patch is obviously structured quite differently)

1 Like

Here is an updated version of this patch that works with recent syntax.

fluid slicer tester.maxpat.zip (33.3 KB)

I just got time to play with this patch, @leafcutterjohn I’m leaving your GenericExperimentalTechnoGenerator™ running for the next hour. I did go blind, but as in Kung Fu Panda: of awesomeness :wink:

thanks @rodrigo.constanzo for the translation.

ps There are plans to add features to the buffvisualiser we have to add markers!

1 Like