Skip to content

Commit

Permalink
Quantile normalize timbre transfer demo.
Browse files Browse the repository at this point in the history 
Better compensates for difference in loudness curves between the source and training data. Also added a crude form of "autotuning". Since we're using Pickle to save and load quantile transforms, we have to freeze the version of scikit-learn as loading breaks between versions.

PiperOrigin-RevId: 314434946
  • Loading branch information
@jesseengel
jesseengel authored and Magenta Team committed Jun 3, 2020
1 parent 2b91b9a commit 04fbd5b
Show file tree
Hide file tree
Showing 4 changed files with 556 additions and 128 deletions.
Loading

0 comments on commit 04fbd5b

Please sign in to comment.