Hi guys,
I've shared today my first experiment with isolating instruments/drums/voice, from a previous CDLC, to give a new CDLC version containing music track without bass, created at start for my sole use and training. ....
As coldrampage said, you could find Machine/Deep Learning tools for this...
I've tried 2 tools to split the music tracks, one online (web site) and one software tool that are using DeepLearning...
- Create stems from any song with ezstems - https://ezstems.com/
- Spleeter - Fast and Efficient Music Source Separation Tool with Pre-Trained Models - https://github.com/deezer/spleeter (free and open-source!)
For the first one, you just have to upload audio file (although there is a size limit), choose number of isolated tracks wanted, launch the process (wait in a priority stack) and download the result.
The second one need to be installed and configured (Windows), wich isn't so easy, but give generally better results.
When using the "5 stems" options we obtain 5 individual splitted tracks:
- bass
- drums
- piano
- vocals
- other
It's all automated and the result is very surprising!!! But your mileage may vary...
Voice and drums can be generally well isolated. [Edit] They can be removed with ususually very good results!
Same for bass when it's not too mixed with other instruments.
But guitar are often found in "other".... but still usable if not much other instruments or sound effects!
Best result is when you want to remove one of the splitted track, for exemple removing bass...
I let you see/listen to the result with bass (almost completly removed) in my new CDLC (from Seneca's previous CDLC - https://ignition4.customsforge.com/cdlc/43889):
Big Girl (You Are Beautiful) - No bass - https://ignition4.customsforge.com/cdlc/67847
I would be interested if you know and practice other ways/tools/software to do the same job.
See U