Skip to content

Pretrained models

blue-fish edited this page Feb 14, 2021 · 10 revisions

Pretrained models come as an archive that contains all three models (speaker encoder, synthesizer, vocoder). The archive comes with the same directory structure as the repo, and you're expected to merge its contents with the root of the repository.


Please ensure the files are extracted to these locations within your local copy of the repository:


Details about model training and audio samples can be found here:

If you're using an older version of the repo code which has the Tensorflow synthesizer, you'll need a different set of pretrained models.