Music-reactive transforms

[user19860331]

I've been following some tutorials recently that make use of the 'Convert audio to keyframes' feature.

I thought it seemed a little too good to be true... and then realised, when trying it out with a song that I like, that the transforms didn't really 'match up' with the timing of the song.

I realised that the keyframes that were higher-up in the graph editor weren't necessarily the sounds that needed the most emphasis in terms of whatever tranform it was (hopefully this all makes sense).

Does this mean that it's generally better to animate things manually when timing to music?

[Ement · Accepted solution]

To get more impactful or obvious use of the audio amplitude, add bass and treble to the audio layer before converting to keyframes. Turn up the bass and subtract treble, or vice-versa. This will help push the low and high frequencies away from each other which should result in what you want.

[user19860331]

Thanks! That has helped a little I think, although I find it still seems like there are too many noises in there to give the desired result.

I think it would be fine for some generic music visualisations, but for what Ive got in my head I'm thinking I might just need to do it manually (I'd probably only need a handful of keyframes and then could just loop it)

[user19860331]

...also I meant to say, I slowed down the audio to 50% speed and realised it was matching up better than I first realised. I think the song was way too fast for me to be able to tell properly originally!

[Mylenium]

As already pointed out, you need to process the audio file to boost the frequencies/ bands you actually need and use those duplicates to generate the animation, not the original audible sound. That applies not only to the keyframe conversion workflow, but also plug-ins liek Trapcode Soundkeys and others that operate based on sound or have audio-related functions. In addition, you often may simply need to enhance your expressions with linear() functions to convert value ranges or use valueAtTime() to account for the differences between perceived sounds and actual on-screen action.

Mylenium