Single Speaker TTS (version 0.2.0)
Full implementation of the architecture using Tensorflows tf.train
API.
This release will be the last release before the entire codebase is rewritten to the new Tensorflow tf.data
API.
The code supports:
- Training
- Evaluation
- Inference
- Exporting (experimental)
- Serving (experimental)
The code comes with examples for training on different datasets:
- rc1-training-ljspeech
- rc1-training-pavoque
- rc1-training-cmu-slt
- rc1-training-blizzard-nancy
Pre-processing of all datasets is implemented using pyfunc
.
Data is fed to the model using tf.placeholder
Tensors that are filled using Tensorflow queues.