Item — HN Reader

Hi, maker of Kicks here.

I don't like the trend where people put language models into everything, contributing to global warming. Samples from sound packs usually have most of the data in the names, it is just very unstructured. I have tons of regexps to figure out different info (like scale) from names.

That said, I'm working on actually analyzing files themselves with Apple Neural engine and pretrained local ML model with some spectral analysis. This will be a huge lift, but this project is a marathon for me until I'm somewhere where the price is well justified

1 comments

As an extra data point, my sample library is an absolute mess with regards to filenames even though 99% of it is from professional packs. I'd love to see some analysis built in, BPM / key / notes etc. For instance, I've got one folder where everything in it is named `STAB 0001` etc. and there's about a thousand of them. Kicks hasn't renamed any of them, but has tagged them with some not so useful tags: `0001`, `1030`, `sample`, `stab` and `stabs` - the first is the latter part of the filename, I've no idea where the `1030` came from but that's on everything from that folder now.

As for the price, it's maybe a little steep for this launch version, but if you get some good results going with the local ML it'll be cheap.

One last thing, it would be amazing if this was a universal app so I could use it on my iPad where I do the vast majority of my music making. Good luck, I like it so far!

Wow, thanks a lot for the feedback. I'd love to keep iterating on tags.

1030 is clearly an attempt to pick up BPM. I try to find a number that's bigger than a simple sample number (like Kick_1 and Kick_2), but have completely forgotten to set an upper limit, will fix that!

What would be your ideal workflow for universal app? Would you like to selectively synchronize your sample library between devices? I've been thinking about it but wasn't sure if someone would use that.