Slicing audio in Drambo based on BPM + Transient detection

aftermast

I’d like to build a sampler that has the ability to slice a whole song based on BPM set to quarter or eighth notes (or even by bar). Ideally, kick notes of the same key could be grouped together and then snares of the same key. This is born out of my Koala workflow, which I love, but is limited in terms of slicing a whole song. I usually chop the audio to 16 pads using auto-chop equal and adjusting the splices as necessary. I really like the variation of chopping two 16 pad loops that are musically the same but from different parts of a recording – but in Koala then that’s half the pads used up.

I’ve been doing a lot of reading on this forum and thought maybe Drambo could help me with this. There are some members (@rs2000 @bcrichards - thank you!) here whose posts and patches have helped me get past just using Flexi and start to make some connections, but, I am far from an expert so I’m reaching out for help at least with the slicing. Could I trigger slices of audio in Flexi Sample using rising offsets calculated from the BPM? Could Transient Detector help with audio that is not perfectly in beat? I know it comes up a lot that some people like to slice longer audio, so maybe there are some ideas out there?

rs2000

Let me first understand what you're aiming at. I see 3 different requests here:
1. Slicing an existing audio file into many individual slices
2. Separating instruments (kicks, snares etc)
3. Detecting transients in live audio

What kind of audio are you looking at and what do you want to have after the magic process?
A while ago you wrote that you've already extracted stems on the command line on your computer - which tools did you use?

I've played with some of these possibilities to auto-sample live audio into a Sampler module so I can play shorter samples, detected from transients, individually.
A few observations regarding full mixes:
Yes, the transient detector will detect rhythmic patterns but this alone won't necessarily give you a nice "bpm grid", you'll often need to do additional processing to get a more useful, cleaner shape.

Unmixing into separate instruments (kick, snare) will be hard without more advanced processing like AI tools which are not available in Drambo. Sure you can do some filtering but the success rate will highly depend on the input material.
Routing audio by simple filtering (e.g. splitting by low, medium, high and low+high frequencies) is fairly easy.

Indeed you can trigger slices at bpm rate (e.g. using the Clock Generator + Counter@16 + [email protected] to generate a rising pitch CV signal fed into Flexi in "Start Slice = Note" mode) but then you need the sliced audio available in Flexi already.
That's why I've preferred to use Sampler: It lets you sample audio snippets based on triggers, e.g. from detected transients, with the sampled length controllable by the trigger signal length. And nothing stops you from using multiple samplers for sampling different classes of detected audio into different samplers so they're easier to use in the end.

aftermast

Thanks @rs2000. I think on number 1 is clear from your response, but I was thinking it could work like the SP Flexi Sampler patch where the the slices are triggered in Drambo based on Offset - and that way the slices don't have to be there.

For #2, I'm not interested in extracting Stems. I was using the command line tool before and it's now built into Koala and other Apps. Basically I was thinking if I know the Kicks are on beat 1 and 3 then to further classify them I'd love to know what notes are in those slices.

For example, Isaac Hayes - Ike's Rap II has been sampled by Portishead, Tricky, Alessia Cara, etc. but mainly one loop. In Koala I can chop that loop into 8 pads and since the first 2 minutes of the song are all based on that same progression, I can chop 7 more versions of the same loop to the remaining pads. With this example, I don's need to know the notes because I know on pad 1 and 9 of each 16 pads will both be a kick at the beginning of the progression. So it would be great, if in Drambo I could have one note that had these samples as layers, but really the slicing would get me most of the way there.

I'm often doing this on an iPhone so maybe that's why Koala works best. Setting slices in Drambo is a pain on an iPhone but just discovered I can run it on a Mac with my same purchase, which is great. I've have ReSlice and tried sEGments a while ago but most samplers where you can set splice points show the entire file and it gets really complex. It would be great if you could move through the file with a set window of a smaller portion of the audio, kinda like SpaceCraft does.

rs2000

@aftermast Well, triggering different parts of a loop is possible without slices but that would make the most sense if your loop follows a straight rhythmical grid.
The waveform editor view in Flexi can be zoomed in with two fingers and the zoomed view can be scrolled with one finger, even on an iPhone, just in case you haven't noticed already.
Triggering "slices" of a plain, unsliced Flexi sample can be done like this:

Scale&Offset processes the pitch signal from MIDI2CV. You can control the section to play by MIDI note and the length by the note duration in the sequencer. Fine-adjust the Cale and Offset knobs to choose which sections of the sample to play. Did you know that horizontal knob tweaks do fine adjustments and vertical moves adjust values in coarse increments?

aftermast

@Rs2000, I got this all set up in Drambo. Well I think I did because it's not working as you described. Could you elaborate on how the Amount / Offset / Range adjustments adjust the sample trigger points? I have played around with it using shorter samples (Dirty Groove 85) and with an Offset of 0.37. This maps the start of the sample to C-1 but if I change the offset I the start trigger is mapped to a different key. Actually it seems that the the triggers are the same no matter what the offset, it's the keys that play them that are changing.

Also, every note I put in the sequencer is playing the start of the sample! Here is how it looks:

rs2000

Is the input of your Scale module connected to the pitch output of MIDI2CV?

aftermast

@rs2000 said:
Is the input of your Scale module connected to the pitch output of MIDI2CV?

Hmm...I noticed that my scale knob was burgundy when it should have been green? I fixed that but it is the same behaviour.

aftermast

I took a totally different approach and learned a lot about Drambo along the way. I think unless a song was recorded with a click track the above method won't work well.

So I'm playing a sample using Flexi into a noise gate and a graphic modulator to mute the sound output between beats so that the flexi that records the sound can record just the rhythm. It is all tied to the BPM set in Drambo.

First you have to record the output of the transients into the 2nd flexi. Detect Transients. Clear the recording and then record the original sample into the 2nd Flexi and this should now have a flexi that has slices on the beat. It works pretty well. Here's an example from Isaac Hayes - Ike's Rap II, which varies from the BPM quite a bit.

Detect Transients in Flexi looks like this:

Using this method:

I am open to any ideas to tweak this but I can't figure out how to attach my Drambo file. The main limitations are the same limitations of the Flexi Sampler:

• You can only record 1 min of audio
• I think if you record the full 1 min of audio, clear the audio to leave just the slice markers, you are not able to record 1 min back into the file
• It would be great if you could just clear the audio in Flexi, leaving the markers and then load a different sample but I couldn't figure out how to do this. You could then have all sorts of templates like "1 minute 130 BPM" just a slice file and then load audio on top of that.

rs2000

@aftermast If you don't mind decoding the Flexi preset file, there's a dirty hack you might want to try.
All you need to do is replace the audio filename and make sure the replaced file has the same length and properties.
The markers should persist.

aftermast

thanks @rs2000! I'm definitely going to have to figure out how to do that.