Sample Slice Auto Detection

Discussion in 'MASCHINE Area' started by DjJascha, Aug 20, 2019.

  1. lwj|localspace

    lwj|localspace NI Product Owner

    Messages:
    361
    I've ran into this as well with complex sounds.. especially if they have bass or wide range of frequencies, the algorhythm doesn't know what to do. But other slicing algorhythms I've tried react pretty similar. Ableton would probably do this too and has worse transient detection sometimes.. where they're all early or late.

    What helps sometimes is to place a DC filter or even a simple high-pass, low-pass. Synths can generate some extreme lows and highs that throw off the detection. When I record my synth I usually do that through a channelstrip VST to just clean up the sound going in. That makes slicing easier as well.
     
    • Informative Informative x 1
  2. b-r@nno

    b-r@nno NI Product Owner

    Messages:
    147
    it would be nice to have an additional parameter one can set before slicing: like a threshold so Maschine know which transients could be ignored and which are the "main ones".

    i mean....usually the highest peak in a waveform is often the point, where "something interesting is happening" and should be detected as a slice...
     
  3. D-One

    D-One Moderator Moderator

    Messages:
    6,681
    Just to clarify... There is, that's exactly how it works, the "sensitivity" value in "detect mode" is a threshold. It does detect peaks/transients as the start of a new slice... The problem here is the OP has sort of an uncommon sample with a lot of dynamic range and a ton of transients... he wants some peaks to be detected at the start but then doesn't want other peaks that are much louder (above threshold) to be detected, at least that is how a interpreted the info given.

    "Detect Mode" is meant for simpler material, it's not any sort of fancy algorithm capable or adapting to high dynamic range, ignoring delays or detecting note/chord changes, for complex waveforms manual mode is required or manually removing auto-detected slices before applying. BPM values dont change anything in the detection unless you use Split mode. This is the same for all samplers I have ever tried, (Including ReCycle) the only one I know that uses smarter detection is ReGroover but then again is meant for drums which are not that hard to deal with in the first place, or Melodyne but calling it a sampler would be quite a stretch.
     
    Last edited: Aug 21, 2019
  4. lwj|localspace

    lwj|localspace NI Product Owner

    Messages:
    361
    Seriously a bit of pre processing can help a lot. Especially when you’re recording a synth. Low frequency rumble, dc offset, lots of high frequency content you don’t need, maybe phasing...

    The detection is never gonna be perfect for all sounds, but if you take out some of those problems first it saves a lot of time.. and sounds better ;)
     
    • Like Like x 1
  5. b-r@nno

    b-r@nno NI Product Owner

    Messages:
    147
    Thanks for your answer d-one.

    Regarding the suggestion of preprocessing:

    Usually i like what i hear when i decide to use a recorded sample and want to slice it.

    Wouldn't it be nice if the suggested preprocessing could be done by using an amount of "unhearable" pre-slicing parameters like HP / LP / BP settings?

    If this could be implented we would have the best solution:

    - algorithm knows "what to listen to" to decide where slicing should happen
    - i wouldnt have to change how my sample sounds

    We could even go further and have other sets of parameters like "Minimum length of slice needs to be x beats / bar / seconds..." or "ignore what you think is worth a slice if its shorter then x beats / bar / seconds...".

    Atm I allways have to adjust the slicing.

    Little example:

    I record the sentence "I Love my maschine". My experience is usually this:

    If sensitivity is too soft, it doesnt even recognise the single words and it makes something like "i love my" "masch" / "ine"

    Sensitivity to high: i / l / o / v.....

    But i cqn live with it. Because the result show me everytime; my voice is not meant to be recorded but "to shut up" *lol*


    Gesendet von meinem TA-1020 mit Tapatalk
     
    • Like Like x 1