keonlee9420/Parallel-Tacotron2

PyTorch Implementation of Google's Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling

47
/ 100
Emerging

Implements differentiable duration modeling through soft-DTW loss and learned upsampling for parallel mel-spectrogram generation, eliminating autoregressive decoding bottlenecks. Integrates fairseq's `LConvBlock` components and adapts FastSpeech2's training pipeline (optimizer, scheduler, FFT blocks) while customizing soft-DTW with CUDA acceleration for GPU efficiency. Supports LJSpeech dataset preprocessing and includes TensorBoard logging; currently optimizing convergence through architectural refinements like sinusoidal positional embeddings and normalized activation patterns.

191 stars. No commits in the last 6 months.

Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 21 / 25

How are scores calculated?

Stars

191

Forks

44

Language

Python

License

MIT

Last pushed

Nov 18, 2021

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/keonlee9420/Parallel-Tacotron2"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.