[Discussion] Can This AI Be Trained for Drum Loop Separating?
- Dominant language
- Python
- Stars
- 28.4k
- Forks
- 3.1k
- PR merge metrics
- No merged PRs in 30d
Description
Hi Everyone, I want to start by saying ive always wanted to play with AI and I am interested in training AI with my own data sets but always thought it would be too hard as I cannot code.
However, now I have managed to get spleeter working in power shell via pip I feel that I am a little step closer to making this happen, with your help of course.
Using the 5 split setting for this example, my idea is this.
I have a vast collection of pristine drum samples I want to use for the data set.
So split 1 would be given, 10,000+ kick drum samples ranging from all different musical style genres. Some would be dry, some would be wet but they would all be kick drums.
Split 2 would be given 10,000+ snare samples and maybe clap samples and rim shots. Again all music genres, dry and wet samples.
Split 3 would be given 10,000 hi hat samples. Closed hh, open hh, drum machine hi hats. Maybe ride cymbals,crashes,splashes,shakers,tambourines, triangles etc should also be put into this set, not sure. Its high end treble percussion.
Split 4 would be given 10,000+ tom samples and other percussion elements falling into the frequency range of toms, like bongos, tablas etc
Split 5 would be given 10,000+ mixed bag "other" percussion. Cowbells, wooden sticks, claves, guiro, vibraslap etc etc
Once trained I want to put a drum loops through the AI to see how it does in splitting the individual drum elements.
Does this sound like a credible idea and would anyone out there want to help me with this?
Contributor guide
Assessment
This issue has not been assessed yet.