Exploring AI Tools for Identifying and Tagging Sounds in Music Production

[denior]

Hello Everyone,

I'm interested in finding AI tools that can analyze sounds and provide tags or identify characteristics, like what type of instrument it is or the specific qualities of the sound. For instance, if I'm trying to recreate a particular sound on my synth and I'm not sure about its nature—whether it's an FM synthesis sound or something else...I'd like a tool that could help by suggesting possible names or characteristics.

Does anyone know of any promising AI tools or software that can do this? The goal isn't to have the AI recreate the sound for me, but to provide insights that can help me learn and become more efficient in sound design and recreation.
Thank you all!

Click to read more...

 

[Sophus]

There are some tools available like the sample browser Sononym, drum sample browsers like Algonaut Atlas or XLN Audio XO.

I'm not sure if these will help you recreating sounds. It's more like sorting a huge amount of samples in a meaningful way.

Besides that, you can try to recreate a lot of sounds simply by trying to follow the basic steps of synth sounds creation. First, which kind of synthesis is used? Are oscillators used (sine, saw, square, noise etc.)? Are multiple oscillators used and combined? Are they slightly detuned to each other? Is granular synthesis used? Or FM synthesis? Next, what's the ADSR envelope? Then, is a filter used and whats the filter envelope? Last, are certain effects used?

You can simply start to replay the original synth sound or melody with a simple sine wave. After you have the correct notes, you can go from there and add other things to your sound like the ADSR and filters, to make bring your sound nearer and nearer to the original.

 

[Barrel Maker]

There are AI tagging services for songs, but I don't know about individual sounds. The AI tagging services for songs are far from perfect, but they can help if you have a catalog of 1M+ songs. A company I worked for used Exactuals RAI and musicube to try to scale the process of manual metatagging.