Quote:
Originally Posted by
beau_mckee
β‘οΈ
Thats great they are acknowledging the source. Hoping someone can answer the other half of this, interested if there is any innovation added to this separation ability in this product?
I think:
it's because they use
their own training set, as mentioned on page 3 of the manual.
In my experience, ReBeat delivers very clean signals.
You can also hear this on the Beatles revival files. Here Peter Jackson's team only used sounds from Beatles recordings as a training set for the AI ββmodeling.
Today, AI engines have become MUCH better than they were back then, but the training sets are much broader.
So the AI ββengines have gotten better, but to satisfy more styles, the separation quality has only gotten marginally better.