Sononym sample management/browser

[danoise]

From what ive gathered the auto-tagging is quite nicely structured, thanks for that! Just wanted to apply tags from scratch for a part of my library.

Then I'd check out the tag panel: the options menu allows you to disable auto-tags.

And yes, documentation is still missing for this feature

[Teeh]

Ah got it, thanks!

[PsYAuM]

Can you reset the trial so as people ccan try out the new version?

[danoise]

Can you reset the trial so as people can try out the new version?

We're not using online activation, but I can offer an extended refund period: three months instead of the usual 30 days. If you should ever need it, just refer to this message

PS: And yeah, it would make sense if every major release offered a new chance to trial the software. Unfortunately we didn't get around to making that change for v1.6.

[Andrei Marchenko]

Somebody else experiencing a little bit glitchy zoom in/out function? It's working but not smooth

[PsYAuM]

What is the secondary sidebar for?
drag panel here to add them?

[danoise]

What is the secondary sidebar for?
drag panel here to add them?

Yep, it's a drop target for whatever panels you want to have visible, always:

[bnz]

You received a 1.6.1 update notification.

[imrae]

Should I be worried about the update tearing up my SSD, or is the high I/O just for the first session or so after updating? Holding off the update until this is a bit clearer...

[danoise]

is the high I/O just for the first session or so after updating?

Yes, once the initial scan is complete, there is hardly any write activity to speak of.

But still, it's being taken seriously. And I managed to reproduce what anoise reported, with almost 10x the data written to disk as being read (the size of the database seems to play a part, but it's too early to draw any conclusions...)

[Digivolt]

Should I be worried about the update tearing up my SSD, or is the high I/O just for the first session or so after updating? Holding off the update until this is a bit clearer...

Just to add some balance, I took the risk and from a 379gb sample folder it wrote a databse file that was 2.7gb on windows 11, haven't had time to jump into sononym yet and start tagging stuff so no idea if that adds to the size and hopefully with updates that size can be reduced further

[danoise]

from a 379gb sample folder it wrote a databse file that was 2.7gb

Ah, but the thing that has some people worried isn't so much the size of the database, as it's the disk activity that went into creating the database.

[Digivolt]

from a 379gb sample folder it wrote a databse file that was 2.7gb

Ah, but the thing that has some people worried isn't so much the size of the database, as it's the disk activity that went into creating the database.

I misunderstood then, thought it was actually writing 10x file sizes although thinking properly that's silly because people would soon be running out of drive space, what exactly are the writes its doing, I would have thought it would just be reads to analyse the files ? And do the writes affect the drive the samples are on or your C: drive where Sononym is installed ?

[danoise]

I would have thought it would just be reads to analyse the files ? And do the writes affect the drive the samples are on or your C: drive where Sononym is installed ?

The only thing that's written to is the database. But for obvious reasons, and many times during indexing.

[waxtrax]

The update is nice, but the performance for very large libraries is still not there unfortunately -- especially for the similarity search (for me personally, this is one of the most useful functions of the software).

I could be completely wrong about this, of course, but the behavior seems to be that audio files are somehow re-scanned during a similarity search (based on the observation of how long it takes). If this is true, I wonder if perhaps an optional separate database could be created that contains all the different properties that each file is searched against for similarity? I understand such a database could end up being huge, but I would definitely trade increased disk usage for increased performance.

Also, is this software multithreaded, at least when it comes to scanning and building the database? I am comparing performance to the latest version of Loopcloud on a Mac M4 Max system with 8TB SSD and 64GB RAM. In a library with around 800k files, Loopcloud will build a database and analyze all of the files in about 30-60 minutes for the first run (which is about 20x faster than Sononym in my experience on my system). Subsequent library additions usually take less than a minute.

Loopcloud takes full advantage of all the cores in the system and processes as fast as possible (while also providing the option to reduce core usage if you need it). Loopcloud will use nearly 100% of my CPU during indexing if I keep it on the fastest-performing setting. Sononym on the same system with 12 Performance cores and 4 Efficient cores only uses about 11% of the CPU.

Then, when doing the AI-based similarity search in Loopcloud, the results are nearly instantaneous. The accuracy is also generally good enough to justify the instant results, in my opinion.

I know there may be some tricks in Sononym to increase the performance a bit, such as partitioning the sample libraries and attempting to limit the search scope, etc, but I don't have to do anything like that with Loopcloud, and it is extremely fast.

I love how Sononym works and how it looks, I just keep wondering if there are some under-the-hood enhancements that could be done (even if they were just optional toggles) to greatly increase the performance for very large libraries.

Looking forward to future updates!