PointZero: Audio Mosaicing with fluid.NMFbuffcross~

Hey All,
Here is my patch that uses fluid.NMFbuffcross~ to perform concatenative sound synthesis in stereo! Sounds great. Thank you FluCoMa team for implementing the “Let It Bee” algorithm… such an absolute game changer.
Feel free to download and tinker with the code, and please share what you make (sound design / patch changes) on this thread! Feedback is welcome.
Big shout out to @tremblap and @weefuzzy for helping get this thing working earlier this summer at the FluCoMa hack day at University of Edinburgh.

Free Download on Gumroad:

Tutorial video:

5 Likes

Ooops, the gumroad link didn’t post right. Lets try again:

2 Likes

Cheeky name :wink:

I will definitely check this out in more detail this evening.

I think my colleagues and I would agree that this should somehow be represented in the flucoma-verse learn platform somehow. We have a space (https://learn.flucoma.org/made-with-flucoma) to fill with wonderful community creations – maybe you’d be up for that in some capacity?

1 Like

Absolutely! :slight_smile: Still some silly bugs to work out tho, all documented in the info pane. Also excited to see how the object evolves beyond this current beta, and will be adding more features if new object params arise.

1 Like

Hi Ben,

Thanks for sharing the video and patch. Very clear and good explanation of the process.

I’m comparing it to results I’m getting from Alex Harkers entrymatcher and from

some more advances ‘misuse’ of Carmine Cella’s ‘orchidea’.

When opening your patch (with Max 8.2, FluComa Beta 2), I’m getting these error messages.

They don’t prevent the patch from working, but might be confusing for beginners.

Screen Shot 2021-11-02 at 8.37.55 AM

After loading a target, similar errors appear, very probably because it is a mono sound.

Screen Shot 2021-11-02 at 8.43.30 AM

This is purely aesthetic: whith long file paths there is a clash with the comment box

Screen Shot 2021-11-02 at 8.43.57 AM

The result is somewhat surprising, as it introduces a lot of ‘percussion’ on a purely vocal sample. But I have to
experiment more with the settings. The corpus are from very similar vocal samples - just a different song. I’m joining a short excerpt of the target and the render result.

Archive.zip (498.1 KB)

1 Like

I am aware of these bugs, thank you!
If anyone has advice for fixing the errors, I’ll make the updates.

Aaaah and the long file path with comment box I have not seen before! Nice catch!

Best,
B

You could check the number of channels with info~ for example, and then, depending on 1 or 2, format the commands for the successive processing chain.

1 Like

This is super cool! Thanks for sharing!