In our recent paper we propose Flowtron: an autoregressive flow-based generative network for text-to-speech synthesis with control over speech variation and style transfer. Flowtron borrows insights from IAF and revamps Tacotron in order to provide high-quality and expressive mel-spectrogram synthesis. Flowtron is optimized by maximizing the likelihood of the training data, which makes training simple and stable. Flowtron learns an invertible mapping of data to a latent space that can be manipulated to control many aspects of speech synthesis (pitch, tone, speech rate, cadence, accent).

Our mean opinion scores (MOS) show that Flowtron matches state-of-the-art TTS models in terms of speech quality. In addition, we provide results on control of speech variation, interpolation between samples and style transfer between speakers seen and unseen during training.

Visit our website for audio samples.

1. NVIDIA GPU + CUDA cuDNN

1. Clone this repo: git clone https://github.com/NVIDIA/flowtron.git
2. CD into this repo: cd flowtron
3. Initialize submodule: git submodule update --init; cd tacotron2; git submodule update --init
4. Install PyTorch
5. Install python requirements or build docker image
   • Install python requirements: pip install -r requirements.txt

1. Update the filelists inside the filelists folder to point to your data
2. python train.py -c config.json -p train_config.output_directory=outdir
3. (OPTIONAL) tensorboard --logdir=outdir/logdir

Training using a pre-trained model can lead to faster convergence. Dataset dependent layers can be ignored

1. (WIP) Download our published Flowtron LJS or Flowtron LibriTTS model
2. python train.py -c config.json -p train_config.ignore_layers=["speaker_embedding.weight"] train_config.checkpoint_path="models/flowtron_ljs.pt"

1. python -m torch.distributed.launch --use_env --nproc_per_node=NUM_GPUS_YOU_HAVE train.py -c config.json -p train_config.output_directory=outdir train_config.fp16=true

1. python inference.py -c config.json -f models/flowtron_ljs.pt -w models/waveglow_256channels_v4.pt -t "It is well know that deep generative models have a deep latent space!" -i 0

WaveGlow Faster than real time Flow-based Generative Network for Speech Synthesis

This implementation uses code from the following repos: Keith Ito, Prem Seetharaman as described in our code.