How to reveal a text with an audio spectrum animation ?

[Ewaren]

Hello,

I'm trying to do something pretty tricky that has to do with matting but I really can't figure out how to properly do it.

Here's the situation : I have a white text layer in my composition (fig 1), and I also have a black background with the "Audio Spectrum" effect on it (fig 2) : this animates a nice dynamic Audio Spectrum that moves with the song I linked it to.

fig 1

fig 2

What I have managed to do so far is place the Audio Spectrum over the text (image 3), and alpha matte my text with the audio spectrum (image 4) : what this does is that pieces of the text appear and disappear depending on whether the spectrum covers them or not at each moment in time.

fig 3

fig 4

However I would like to do a bit more : I want the pieces of text to stay revealed once the waveform has gone through them at least once, so that they don't disappear when the frequency of the music changes, and so that eventually if the music covers enough frequencies all of my text will be revealed. To sum it up, what I want to do is sort of a "persistent" reveal effect once the waveform has gone through a certain piece of text.

I can't seem to find a way of doing that easily, because with the simple matting options the pieces of text will obviously disappear when the waveform leaves their area, and then reappear if the waveform's frequency matches the area again.

Does anyone have a solution to create this ?

Thanks in advance

[Szalam]

You could possibly use the Echo effect on the Audio Spectrum layer that's matted by the text. If you have it set to screen and crank some of the settings, it might do the trick.

[Mylenium · Accepted solution]

You need to pre-compose a duplicate of the wave and then mangle it with temporal effects like Posterize Time, Echo and possibly native time-remapping. It's still going to require lots of manual fiddling, though. There simply is no way to predict/ control how the audio is sampled and what kind of pattern it will produce and combined with temporal sub-sampling this may never give the result you need at any given point. In fact it might be smart to just pre-render the entire spectrum and reimport it to have more control.

Mylenium