You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Following a question I received through an email, here is a small script to control pitch. It can be placed and ran wherever in the working directory and assumes you downloaded models using the src/utilities/download_checkpoints.py script.
Pitch is extracted from audio using the analyzer and can then be modified before being fed to the synthesizer. Here 2 options are explored, pitching down and monotonic pitch, but other manipulations can be thought of. One could for instance use a transcriber to get precise timestamp for each word and then apply a specific pitch at each location to totally remodel the pitch sequence.
Following a question I received through an email, here is a small script to control pitch. It can be placed and ran wherever in the working directory and assumes you downloaded models using the
src/utilities/download_checkpoints.py
script.Pitch is extracted from audio using the analyzer and can then be modified before being fed to the synthesizer. Here 2 options are explored, pitching down and monotonic pitch, but other manipulations can be thought of. One could for instance use a transcriber to get precise timestamp for each word and then apply a specific pitch at each location to totally remodel the pitch sequence.
The text was updated successfully, but these errors were encountered: