I’m trying to match the EQ of two audio tracks, one is a clip from an episode of Lost and the other is something I recorded.
I used the plot spectrum tool and found that was basically a slope from the top left down to 15000hz. My clip starts off the same, dips at 1000hz, comes up swinging around 5000hz then drops again.
I can use the equalisation tool to match the spectrum, but it takes forever.
Is there an easier way, some sort of plugin that can spot the difference between the two clips and then alter one of them to match the other?
If not can someone recommend a way to measure this? At the moment I am just taking shots in the dark, checking which frequency is low, moving the eq slider, checking what happened, undoing, modifying the eq and redoing.
Is there a more scientific way to do this?
Certainly. When you do the analysis, pull the window as large as your screen will allow. See the numbers on the left? Those are dB values of the loudness of the energy. Frequency along the bottom.
Do the same thing with the Equalization tool. Pull the window really large. Those numbers on the left are the same dB values and the bottom is the same frequency listing
Let’s see, 4 KHz needs a 3dB boost. 7 KHz is 8 dB too hot.
You can cheat and print the analysis and compare that with the active equalization window, or you can switch rapidly between them, or, if your computer has two screens…etc.
This will only get you in the ballpark. Chances of you getting a really good match are pretty slim, but it will be an interesting exercise.
As a related task, I always wanted to do a spectrum analysis of my truck cab and equalize CD music so it comes out flat instead of the resonant dips and peaks that I know are there.
After reading the response from koz, and completely agreeing with what he says, I would totally contradict him and say “no, there is no no more scientific method”. I have come across “spectrum profiling plug-ins” and they work to some degree, but at the end of the day, the algorithms that specify how a song is perceived are generic approximations.
Being able to identify frequencies and dynamics comes with practice. There are tools (such as the spectrum display) that can help, but in the end it comes down to listening, and making your own musical decision.
I don’t think i’s going to work any more than you do, but if you want actual numbers for tool setting translation, that’s how to do it.
This process will not affect dynamic compression, presence, or echo, so at best equalization will only get you about half way there.
There is a joke on one of the video forums about a programmer there writing the “Professional Audio” filter. No matter how you shot your piece, just apply the filter…
There could be something. It sounds like you want an equalizer settings that make the average spectrum of both
tracks exactly the same.
If this would be the question (makes good sens only for two recordings of the signal, or for 1000 hours recording of same type),
then this is completely scientific (mathematical) question.
I remember DSP having white noise generator and an analyzer for analysis,
and tool that converted spectrum-output to built in (digital) equalizer setting.
In one click you can obtain equalizer setting for given room.
This is same situation as above, with one track being white noice, and the other track the noise changed by the room.
So this is only matter of manipulation with numbers.
At the moment in past, I though the manipulation was just prepending minus to the numbers.
(If it is +2 dB, make it -2 dB by equalizer).
But since a slider of an EQ-band influences the neighbouring bands,
the inverse matrix should be involved. (I mean given the values of curves of EQ, the inverse matrix to matrix of that.)
If not done yet, should be possible to do for your favorite EQ (assuming it is linear) by measuring it
against sin signal and doing inverse matrix computation in Excel or whatever software (This part is not numerically sensitive at all). — This is the EQ part of thing, and of course, the for the usual EQ that all should be known even without measuring.
I suspect given that you are writing for a free software program, that the poster doesn’t have access to the original live recording or any information about the performance.
Also given the cost of the software, I’ll bet the poster is trying to produce a top-quality, professional audio track without paying for the engineers artists, software, or equipment.
Just a guess.
<<>>
I like reading your posts. You have a slightly unusual word presentation.
(And why should he? Did not ask to match anythign to original, just to tracks one to the other.)
Also given the cost of the software, I’ll bet the poster is trying to produce a top-quality, professional audio track without paying for the engineers artists, software, or equipment.
Just a guess.
<<>>
I like reading your posts. You have a slightly unusual word presentation.
Koz
This time I understand you less than a bit, and perhaps this time you understand me no more that a bit, too.
I hope that otherwise my post are sometimes usefull.
I basically wanted to say it IS ‘scientific’ problem if we only have to make the two curves identical. And can be done
by a script. Needs only to know the two curves, the parameters of band EQ a do the matrix inversion.
The Nyquists scripts around are about that complexity or higher, so there might exists some that could be used, in principle.
The question was about matching the spectrum. (I do not know IF it makes it top quality.)
Some corrections of myself. My “DPS” was a digital hardware, used as motivation.
I did not propose to buy anything. I supposed everything to be a (nuquist) plugin.
Assuming band EQ, not parametric one (This was what I meant by ‘linearity’).
–
Onother way arround: Is there a script that computes spectrum of (microfone) white noice recording
and proposes band-EQ settings for the room where recording was taken? This is about 85% of solving the problem.
(97%, if you have the author of the script nearby.)
AND back to the top-quality: Yes, probably all gets done completely other way if the goal is about how it sounds. And it is.