An opensource speech-to-text software written in tensorflow.

Python 3 is required.

Currently speechT is based on the Wav2Letter paper and the CTC loss function.

The speech corpus from https://www.openslr.org/12/ is automatically downloaded.
Note: The corpus is about 30GB!

The data must be preprocessed before training

python3 preprocess.py

Then, to run the training, execute

python3 train.py

Important flags
--data_dir to specify the data directory to download speech corpus to (defaults to ./data/)
--train_dir to specify the train directory to save checkpoints and vocabulary to (defaults to ./train/)

Not yet implemented.

Not yet implemented.