dannyflint

I want to analyse a waveform, take the amplitude data and apply that envelope to another soundfile of the same length. If I take an envelope follower and map it to the gain of another sound it's far too clunky. I'm looking for an accurate offline solution.

Here is a link to a project file with my thoughts etc.

I want to analyse a waveform, take the amplitude data and apply that envelope to another soundfile of the same length. If I take an envelope follower and map it to the gain of another sound it's far too clunky. I'm looking for an accurate offline solution.
Here is a link to a project file with my thoughts etc.
Any suggestions would be welcomed

offline-amplitude-analysis

not sure what would be different when doing the same thing offline ... but what you should try is to use different times for up and down, just like the attack and release parameter of a compressor or peak limiter, because you probably want to have a short attack.

oh, and dont only measure power history, measuring peak is far more interesting.

1. I drag and drop an .aif (for the sake of explanation lets say it's a 1 second soundfile where the wave looks like two circles side by side)

2. The average amplitude of each group of samples (bin?) is analysed. Or even every sample (perhaps in Gen~ or something) for maximum accuracy.

3. I take another file (for example a 1 second brick of white noise) and apply the amplitude data to this so this wave has the exact volume characteristics.

1. I drag and drop an .aif    (for the sake of explanation lets say it's a 1 second soundfile where the wave looks like two circles side by side)
2. The average amplitude of each group of samples (bin?) is analysed. Or even every sample (perhaps in Gen~ or something) for maximum    accuracy.
3. I take another file (for example a 1 second brick of white noise) and apply the amplitude data to this so this wave has the exact volume characteristics.


If a 1 second sound (at 44.1khz sample rate)is broken down into a list of 44100 floats from -1 to +1, then if you applied that set of values to a different sound it would become the original sound?

yes, of course. samples represent amplitude, first of all.

if you add A to B the two sounds will be summed, if you multiply them with each other thats called amplitude modulation and will probably end up with terrible noise. ;)

measuring peak and/or power must happen in windows of at least a few samples to make sense.

what you can measure and reproduce is often not exactly what you would expect to hear, though.

analyzing the "envelope" of a thunder sound and applying it to a recorded guitar phrase sound can not be archieved by a simple mathematical solution.

you will sooner or later find sound samples where you think they have a hard attack, but in fact the attack has less bass than the rest, and so in the end might have _less energy than the first 40 milliseconds. thats why i say peaks is more important than power.

the whole thing is more a trial and error job until you get the parameter ranges which fits to your taste.

meter~ measures peaks, average~ measures power, if the control signal (the "envelope") is getting too harsh, you can smooth it with something like slide 25 5000.

to get an envelope style control signal based on peaks, first do abs~ before getting the peaks.

when using average~ to get the power history of A with a window of 80 milliseconds, you must delay B for 40 milliseonds to copy the envelope over at the right time.

what you can measure and reproduce is  often not exactly what you would expect to hear, though.

analyzing the "envelope" of a thunder sound and applying it to  a recorded guitar phrase sound can not be archieved by a simple mathematical solution. 

meter~ measures peaks, average~ measures power,  if the control signal (the "envelope") is getting too harsh, you can smooth it with something like slide 25 5000.

when using average~ to get the power history of A with a  window of 80 milliseconds, you must delay B for 40 milliseonds to copy the envelope over at the right time.


Offline Amplitude analysis