A TensorFlow Implementation of DC-TTS: yet another text-to-speech model

fork of kyubong's TensorFlow Implementation of DC-TTS

from Efficiently Trainable Text-to-Speech System Based on Deep Convolutional Networks with Guided Attention.

Contact

fork maintained by derek@atlaslabs.ai

Requirements

• NumPy >= 1.11.1
• TensorFlow >= 1.12.0
• librosa
• tqdm
• matplotlib
• scipy

Changes from Kyubong

• data loading for python3 (remove codecs and other fixes as seen in issue #11)
• absolute paths to data (allow preprocessed files to be written to; read from specified absolute path)
• argparse for specifying task number and GPU (train.py and synthesize.py)
• allow custom sentences, output directory when generating, and output spectrograms as images

Example

rice is often served in round bowls

text2mel: ~50K steps; SSRN ~15K steps

text2mel: ~70K steps; SSRN ~20K steps

Data

LJSpeech

1. download LJ Speech Dataset 1.1
2. rename metadata.csv to transcript.csv (will update later...?)
3. add absolute path to corpus do hyperparams.py as data=<path>

Training

1. check corpus path in hyperparams.py
2. run python(3.6) prepo.py to create mel and mag subdirectories (this speeds up training a lot)
3. run python(3.6) train.py -n 1 to train text2mel. use -g flag to specify GPU id.
4. run python(3.6) train.py -n 2 to train SSRN (mel upscaling). use -g flag to specify GPU id.
5. note: both training can be run together using separate GPUs

Monitoring

tensorboard can be used to monitor training by specifying the log directory (for LJS, LJ01-1 and LJ01-2), e.g.:

tensorboard --logdir=/home/derek/PythonProjects/dc_tts/logs/LJ01-1 --port=9999

Synthesis

1. run python(3.6) synthesize.py, specifying GPU with -g, file with -f (format like harvard_sentences.txt), output dir with -o

the default file and output are specified in hyperparams.py

the spectrograms are also written as images

(from Kyubong master) Notes

• The paper didn't mention normalization, but without normalization I couldn't get it to work. So I added layer normalization.
• The paper fixed the learning rate to 0.001, but it didn't work for me. So I decayed it.
• I tried to train Text2Mel and SSRN simultaneously, but it didn't work. I guess separating those two networks mitigates the burden of training.
• The authors claimed that the model can be trained within a day, but unfortunately the luck was not mine. However obviously this is much fater than Tacotron as it uses only convolution layers.
• Thanks to the guided attention, the attention plot looks monotonic almost from the beginning. I guess this seems to hold the aligment tight so it won't lose track.
• The paper didn't mention dropouts. I applied them as I believe it helps for regularization.
• Check also other TTS models such as Tacotron and Deep Voice 3.