I love the sound of real drummers!

Freebird

I've been listening to Topper Headon, Stephen Morris, and Boris Williams this evening, and came to the conclusion that they are the reason why their bands sounded so great. It's like having a brain massage when you listen to unquantized music. I can't pick a winner though, as they were all fantastic drummers.

Moving forward, I can see e-drums, along with decent software and a real drummer as being a good compromise for my recording needs. It's the timing imperfections and movment that gives life to music!

WiresDreamDisasters

This might be interesting, or not.

The technology used to detect transient content within audio has been around for a long time now. Improvements keep coming however. The latest hotty on the scene is Harmonic Percussive Source Separation. You can hear this in a few products out there, Izotope RX7 has it built into their "music rebalance" tool, and Regroover by Accusonics is also an example of this.

In the old days you'd calculate the onsets of drum hits using various techniques like calculating the zero crossing rate for each frame of audio. The higher the zero crossing rate, the more chance of there being a snare drum present. You'd also compute an energy novelty function for each frame of audio; the largest jumps in energy corresponding to the onsets of notes or drum hits.

So when you run a "detect transients" function in your DAW, and it doesn't do it 100% then this is why. The algorithms being used just aren't accurate enough a lot of the time. Usually requiring manual tweaks.

In short - the tools we're given aren't yet good enough to automate the process of implying the timing imperfections and "choices" that real drummers make, onto a piece of audio or midi. Adding gaussian noise to the start time of a midi note is not what real drummers do.

pmbomb

https://www.youtube.com/watch?time_continue=2&v=hT4fFolyZYU

thecolourbox

On the other hand, I rather like pretend drums but more so when whoever has done them is not trying to sound like a real drummer. I find whilst great drummers are truly brilliant, they are few and far between so working with average drummers is not with the effort for the small potential benefits

(Says the anti social musician who last played alongside anybody else four years ago)

danishbacon

I've been trying off and on for about two years to come up with some interesting beats using a groovebox. Far harder than it seems and I haven't managed to wrap my head around any more complex patterns when using step sequencers, feels like I'm constantly trying to catch up to the Maschine.

Then a Roland eKit came into my life and everything changed. 

Freebird

thecolourbox said:

On the other hand, I rather like pretend drums but more so when whoever has done them is not trying to sound like a real drummer. I find whilst great drummers are truly brilliant, they are few and far between so working with average drummers is not with the effort for the small potential benefits

(Says the anti social musician who last played alongside anybody else four years ago)

I'm lucky enought to know a pretty good drummer, but it's always been easier to go with the midi loops. I started using them as a metronome of sorts and it turned into a habit, but I think it's time to start replacing them with some original parts. Everything else is done by hand, so why not the drums?  I have previously hooked up my laptop to his electronic Roland kit and grabbed a few midi takes, which worked out ok, so I may try it with some live bass next time.

Freebird

WiresDreamDisasters said:

This might be interesting, or not.

The technology used to detect transient content within audio has been around for a long time now. Improvements keep coming however. The latest hotty on the scene is Harmonic Percussive Source Separation. You can hear this in a few products out there, Izotope RX7 has it built into their "music rebalance" tool, and Regroover by Accusonics is also an example of this.

In the old days you'd calculate the onsets of drum hits using various techniques like calculating the zero crossing rate for each frame of audio. The higher the zero crossing rate, the more chance of there being a snare drum present. You'd also compute an energy novelty function for each frame of audio; the largest jumps in energy corresponding to the onsets of notes or drum hits.

So when you run a "detect transients" function in your DAW, and it doesn't do it 100% then this is why. The algorithms being used just aren't accurate enough a lot of the time. Usually requiring manual tweaks.

In short - the tools we're given aren't yet good enough to automate the process of implying the timing imperfections and "choices" that real drummers make, onto a piece of audio or midi. Adding gaussian noise to the start time of a midi note is not what real drummers do.

Very interesting, and there has got to be a way of getting rid of the plodding effect that occurs when you the cycle through repeated drum loops. What do you think about using an electronic kit to capture a whole performance, so that every bar is different, and then running the midi output through VST drum software to fine-tune the sound? Dealing with an analogue kit in the home is probably not worth the bother, and recording studios are better suited for that sort of thing anyway.

PolarityMan

So my take on this is that velocity variation is the absolutely critical thing to get right first. You can go a long way with totally programmed drums or edited drums as long as you get the velocities to work. Something as simple as automating your hi-hat velocities to replicate the "1234" accent pattern of a real drummer can make a night and day difference.

After that the next most important thing it to match the macro timing variations. So that means tempo map up a bpm or 2  if you want the choruses to sound a bit more exciting for example.

Next get the big timing variations at the micro level sorted out. If you analyse a real drummer the largest type of timing "discrepancy" is swing. You can set your grid when programming to a swung grid and get the swing set without any judgement and still get a pretty good outcome.

Only when I've sorted out all that stuff do I start caring about the really tiny things, and you'd be surprised how far getting those 3 things will take you. I rarely bother with anything further for demo purposes but next up would be speeding up slightly during rolls or moving the snare back beat slightly before or after the beat to simulate a pushing or relaxed feel. If you;'ve got the rest right this is really subtle though.

If you enjoy rock music from the last 5-10 years then I would also be willing to bet that at least some of the performances you like have been edited and sample reinforced even if they haven't been totally strictly quantized. 

Freebird

PolarityMan said:

So my take on this is that velocity variation is the absolutely critical thing to get right first. You can go a long way with totally programmed drums or edited drums as long as you get the velocities to work. Something as simple as automating your hi-hat velocities to replicate the "1234" accent pattern of a real drummer can make a night and day difference.

After that the next most important thing it to match the macro timing variations. So that means tempo map up a bpm or 2  if you want the choruses to sound a bit more exciting for example.

Next get the big timing variations at the micro level sorted out. If you analyse a real drummer the largest type of timing "discrepancy" is swing. You can set your grid when programming to a swung grid and get the swing set without any judgement and still get a pretty good outcome.

Only when I've sorted out all that stuff do I start caring about the really tiny things, and you'd be surprised how far getting those 3 things will take you. I rarely bother with anything further for demo purposes but next up would be speeding up slightly during rolls or moving the snare back beat slightly before or after the beat to simulate a pushing or relaxed feel. If you;'ve got the rest right this is really subtle though.

If you enjoy rock music from the last 5-10 years then I would also be willing to bet that at least some of the performances you like have been edited and sample reinforced even if they haven't been totally strictly quantized. 

You should make a video demonstrating your workflow, as it sounds pretty useful  

The stuff I was listening to last night was from the 80's, so none of that quantizing voodoo existed back then.