uberduck-ai / uberduck-ml-dev

ML models for Uberduck
Apache License 2.0
377 stars 61 forks source link

Compute pitches #122

Closed sjkoelle closed 1 year ago

sjkoelle commented 1 year ago

Enable computation of pitches and writing to disk using the data loader. In general, there is some ambiguity about different parameter types (e.g. torchmoji, speaker encoder, pitch, text, mel) are treated by the data loader, and so this refactor tries to clean this up and make it more generic, for example to solve the problem of loading audio files without texts.