Pretrained models

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.

Initial commit (latest release) [Google drive]

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

encoder\saved_models\pretrained.pt
synthesizer\saved_models\logs-pretrained\taco_pretrained\checkpoint
synthesizer\saved_models\logs-pretrained\taco_pretrained\tacotron_model.ckpt-278000.data-00000-of-00001
synthesizer\saved_models\logs-pretrained\taco_pretrained\tacotron_model.ckpt-278000.index
synthesizer\saved_models\logs-pretrained\taco_pretrained\tacotron_model.ckpt-278000.meta
vocoder\saved_models\pretrained\pretrained.pt

Here is some information about the models. For reference, the GPUs used for training are GTX 1080 Ti.

• Encoder: trained 1.56M steps (20 days with a single GPU) with a batch size of 64
• Synthesizer: trained 278k steps (1 week with 4 GPUs) with a batch size of 144
• Vocoder: trained 428k steps (4 days with a single GPU) with a batch size of 100