Skip to main content

Sample folders take significantly longer to load -> analysis takes too long

Bugs - Rando

Compared to the previous release, sample folders are taking considerably longer to scan and load in this beta. Previously, adding folders completed within a few minutes. Currently, the loading process is still ongoing while typing this report.

Posted by Beautiful7 ·
Status: Completed8 comments

Log in to comment and vote

Comments8

  • statmartonline

    •

    Apr 2

    Confirming that latest beta (#6) is still very slow with scanning.

  • Milkman

    •

    Feb 24

    So right now, I have a single folder of about 20,000 samples that I have set Rando2 to scan. 30 minutes later, this is at 10% completed. CPU use seems relatively low (under 6%), with also relatively low disk I/O. Im seeing brief spikes on my nvme, but usually under 15%. RAM use has been between 300-1000MB, usually on the lower side.

  • Milkman

    •

    Feb 24

    Im an extremely edge case, where I regularly have Rando2 and other sample engines to index around 1.2 million samples, mostly wav, flac, some mp3 and ogg. Rando2 (non beta) takes around 25-30 mins to do a FULL rescan of all locations. Rando2(beta) took 6 hours++, then I killed it and it never completed.

    Dennis talked to me via email about this and explained that I should try a smaller list of files for testing, so Ill be doing that and replying with what happens soon.

  • Andyman

    •

    Feb 22

    I’m on Win10, Fender Studio Pro 8, and the sample scan takes absolutely forever. It doesn’t seem to matter if I scan a small folders or a huge set.

    The scan is very fast up to around 20%, then it slows down to an unusable crawl. I’m basically unable to scan my samples.

    No problems in scanning with the production version.

  • Dennis De Laat

    Team•

    Feb 18

    @Beautiful7 are you on mac or pc?

    • Beautiful7

      •

      Feb 18

      Mac M2 Max

  • Dennis De Laat

    Team•

    Jan 27

    Yes that is due to the audio classification. We will make it work so it is on demand.

    • njlang

      •

      Feb 25

      Just in case you aren’t already doing something like this - perhaps it could help decrease analysis time for longer audio loops by scanning smaller chunks of it?

      I recently was checking out the essentia.js online models and their demos state

      “We select 65% of the audio, evenly spaced throughout the track, for analysis to ensure efficient processing while maintaining accuracy.”