waveglow
A flow-based generative network for fast, high-quality speech synthesis from mel-spectrograms.
What is waveglow?
WaveGlow is a flow-based generative neural network for speech synthesis that converts mel-spectrograms into high-quality audio. It solves the problem of slow autoregressive audio generation by providing fast, efficient synthesis while maintaining audio quality comparable to state-of-the-art WaveNet implementations. The model uses a single network architecture trained with maximum likelihood estimation for simplicity and stability.
AI researchers and engineers working on speech synthesis, text-to-speech systems, and generative audio models who need fast inference without sacrificing quality.
Developers choose WaveGlow for its combination of fast inference speeds (1200 kHz on V100 GPUs) and high audio quality, achieved through an elegant flow-based architecture that eliminates the complexity of autoregressive models while maintaining competitive Mean Opinion Scores.
Overview
A Flow-based Generative Network for Speech Synthesis
Use Cases
Best For
- Real-time speech synthesis applications requiring low latency
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
