Sononym sample management/browser

[Gamma-UT]

That tool doesn't exist today. If you're more focused on organisation than search, as I say, I think what you want is Samplism but maybe with some ML front-end that auto-tags or names samples based on what it thinks it finds (which doesn't exist except maybe in research). Editing tags in Samplism is very easy. It could do with some refinements to the UI for that but I've found it to be a lot quicker than the old stalwart AudioFinder in that department. I'm also not that bothered about categorisation overall - just as long as it comes up with enough sounds in the right ballpark on a search.

With a similarity-based search like Sononym the big problem is whether the ML model reflects what humans hear. Fixing the tagging won't help in that department - because it will probably make the same mistake on a rescan. I don't know if there is some mechanism by which users can help with this but what makes more sense over time is to have some type of system by which users can report glaring mis-categorisations and the data from those samples is used to help refine the core ML model.

[aMUSEd]

How does Samplism auto tag, what is it based on?

[woodsdenis]

How does Samplism auto tag, what is it based on?

It does it simply by the name which in most cases works as most samples are named correctly. It means it’s very fast and easy to re tag.

[pinki]

I am delighted with Sononym. The ability to filter results of a search with e.g pitch range or noisiness is a game changer in my opinion. So much creative possibility.

[danoise]

Hey, nice feedback people!

Indeed, Sononym makes mistakes when categorizing content, but as it was pointed out, any "obviously" wrong results should be rooted out. So, the didgeridoo might be a recognized as a bass instrument, and compared to other bassy sounds this is not an wrong match per se. What's important here and now (v1.0) is that searching for that didgeridoo sample will yield similar-sounding results.

When it comes to tagging, this is planned too. We are aiming for a "best of both worlds" approach that allows you to use the machine-learning as a starting point but with additional tags on top. Bringing tagging into the browser/table isn't too hard, but it will of course get really interesting if user tags manage to influence similarity search results as well

[aMUSEd]

That would be ideal thanks and enough to hold me off from Samplism for now

[DrGonzo]

How are you guys feeling about Sononym now when it has been out for a while? You still using it - or have you moved on to something else?

/C

[dandezebra]

I really want this as my demo is over! An ability to override / tag multiple files at once would be fantastic (like changing the genre of multiple bands or albums in iTunes). I think they said that something is coming like that!

[Kalamata Kid]

Is Sononym similar to the free ADSR Sample Manager
https://www.adsrsounds.com/product/soft ... SK-Rekkerd#

[yellowmix]

I ditched ADSR after I demoed and bought Sononym. It could ingest my sample library (~250k) in about a day, database is about 4GB on an SSD. ADSR's never finished, and I let it run for 3 days. Sononym is rock stable, ADSR's crashed quite a bit. ADSR's added BPM to too many non-loop samples and it interferes with searching. ADSR's is based more on the filename, Sononym doesn't seem to care about that much, since it is analysing the actual sound.

Of course it's not perfect but the main category is usually reasonable, and it will multi- and grey tag when it's not entirely sure. Wavetables, for example, are usually classified as Bass. Impact foley can be classified as various cymbals, snares, etc.. instead of breaks & smashes. But there are so many other thoughtful measures. There's class (oneshot or loop), pitch, BPM (only for loops and it's very good at detecting them), brightness, harmonicity, noiseness, rms, and you can sort the list by any of those. There are so many filter categories, including the aforementioned, and file info, and you can limit it by range. Then there's the similarity search, and you can prioritize the aspects: overall, spectrum, timbre, pitch, amplitude.

Sample playback is generally fast. You can browse by database or by file system. File system browse isn't analysed, so other sample browsers generally beat it. I like Resonic Pro; it's extremely fast and can loop sections. Anyway, nothing beats Sononym at classifying and searching for similarity. I am waiting for rx2 support, manual categorizing/correction, and it could use A-B looping, but otherwise I am very satisfied. Demo it and see if it works for your workflow.

I was also demoing Algonaut Atlas, but it also wants to be a drum machine, so I'm waiting for them to improve the classification and search aspects. I like the map concept but it's not too good with large libraries. Sononym's search workflow is faster and better in general. I also like the play random sample when I've got the entire database (I have smaller databases to narrow the search down), it's inspiring.

But since ADSR's is free I can't complain too much. It's wonderful for what it can do and they're open to feedback and are working on it. Competition's good and it'll be interesting to see how this develops.

[Kalamata Kid]

Yellowmix,

Thanks for the detailed response.
https://www.sononym.net/
Does it go on sale?

[DrGonzo]

Bought it. Now for some late-night scanning

@Kalamata Kid. Last year it was 50% off during Black Friday. But no sale during the winter holidays. So I would say yes, but rarely.

/C

[dandezebra]

I would purchase right now if there were a sale! Ha!

[bundoo]

If I add samples to my hd will sononym automatically pick them up and categorise them or do i need to redo the whole file sweep thing?

[DrGonzo]

You need to do a refresh. But that just deals with the changes.

/C