Pajlada TTS

Stripped down version of ForwardTacotron (https://github.com/as-ideas/ForwardTacotron) with pretrained weights for Pajlada's (https://github.com/pajlada) voice.

⚙️ Installation

Make sure you have:

• Python >= 3.7

Install espeak as phonemizer backend, (for macOS use brew):

sudo apt-get install espeak

For windows you can use the installer (https://github.com/espeak-ng/espeak-ng/releases), then make sure that in your path PHONEMIZER_ESPEAK_PATH points to espeak-ng.exe

Then install the rest with pip:

pip install -r requirements.txt

If you aren't going to use CUDA, you can install the smaller CPU only torch version from https://pytorch.org/get-started/locally/

Get the pretrained weights and extract the checkpoints folder to the root where gen_tacotron.py is:

https://drive.google.com/file/d/13I_x2bU6rXTqqIe9Lj8lZyS5OaW5TId-/view

❓ Examples

Generate all sentences from sentences.txt:

python gen_tacotron.py --config config.yaml wavernn

Generate given sentence with lower quality Griffin-Lim vocoder, forcing CPU use:

python gen_tacotron.py --config config.yaml --input_text "Freedom means not having a master." --cpu griffinlim 

Autoregressive models can get stuck in loops if you try to generate too tricky or large sentences. Ending all sentences with a full stop helps.

Acknowlegements

• https://github.com/pajlada
• https://github.com/as-ideas/ForwardTacotron
• https://github.com/keithito/tacotron
• https://github.com/fatchord/WaveRNN

Copyright

See LICENSE for details.