How do I find tempo in an audio track with very fuzzy beat limits

I have an audio track, basicly a drone-like noise with a subtle beat detectable by the ear, but invisible in the wave editor (at least with my newbie skills). I want Cubase somehow to figure out from the “noise” the beats and add a midi “signal” for each of them and set the BPM tempo along the track based on those beats. I have tried the Hitpoints, AudioWarps etc, but threashold/intensity do not identify the beats. I was thinking that somehow maybe the wave editor could show the wave only for the frequencies for which these beats occur. But I have not succeeded in finding anything bringing me closer.

What I have tried starting out on is as follows:
Manually tap an instrument track when simultaneously listening the wave track. However, this is very error prone and time consuming as the “drone” goes on for about 10min. And a subsequent problem rises, as the manually tapped beats must be synchronized better to the wave track.


The end result I want to achieve is to synchronize an external device (some eurorack stuff) based on the subtle drone beats.

Have you tried Tempo Detection? (Project>Tempo Detection…)

I have, but the beats are indistinguishable for Cubase from the drone “noise”.

Hard to say without hearing what the audio sounds like.
You could try to perform some heavy EQ-ing to try and bring out the percussive parts, perhaps in combination with an expander or transient shaper to help with tempo and hitpoint detection.
One step further would be to use a tool like Spectralayers.

1 Like

The recommendation to use narrow, strong, EQ, to extract some kind of signal that could be detected, is what I would try, too.

If that doesn’t work out, then perhaps the better option is to figure out what generates the beat upstream, and try to record that input signal (some LFO or whatnot?) to derive the beat from that?

Computers are great and all, but there are things they still can’t do. Sometimes, you just have to go through the entire track and manually place the markers. This is especially true when the beats aren’t actually on a regular grid. (If they are regular, you just need an accurate tempo indicator, and a regular first hit.)

Is the tempo constant? If so, the easiest thing is to set the track with the audio to linear time base, set a metronome going, and adjust the project tempo (and the start of the audio) until you have a match.

If not, and you can’t alter the signal to accentuate the rhythm, it’s going to be a manual exercise.

Or you could combine the two and set up a tempo track, and adjust that if the beat isn’t quite constant.

Agree this is the best approach to initially try.

Here’s how to do that if the OP (or others) don’t know.

  • Use one of the built in Parametric II EQs
  • Set the EQ Frequency to the max (or min)
  • Set the EQ Gain to approx. 10 to 15dB (or more if needed)
  • Set the Q to the max to get the narrowest band
  • Slowly sweep the Frequency control downward
  • At some point(s) the sound you are looking for will jump out
  • You could just use this as is for the Tempo Detection, but I’d Render it to preserve the EQ’ed Audio

I have thought of this, but left it temporarily because I thought it would be a too time consuminig manual excercise. But, yes, it’s on the list of alternatives.

Yeah, the thought struck me that SpectrealLayers could have something to offer here. I will have a look into that tomorrow together with the ideas from @raino below.

Seems like a possible way to try, I will see into it tomorrow together with the Spectrallayers suggestion from @mlindeb above.

I went on with the idea from @jwatte and @raino . I think that’s the best way. It showed up that the beats (or more sounding like a fuzzy 100Hz pump) are very irregular (“off beat”), but on an average, there is a regularity as far as I can hear and see :smirk: Still, I think there remains a lot of parameter tweeking and manual work to get it the way I want. Anyhow, thanks to all of you pointing me in what I suspect to be the right direction.

(Of course, if any of you is interested and have the time, I would be very curious about how you profs would go along doing it. Hint me, and I will put the file on my Dropbox.)