WOLONet: Wave Outlooker for Efficient and High Fidelity Speech Synthesis

TL;DR

WOLONet is a novel lightweight neural vocoder inspired by Vision Outlooker, achieving high-quality speech synthesis with fewer parameters than state-of-the-art models like HiFiGAN and UnivNet.

Contribution

Introduces WOLONet, a lightweight neural vocoder with a novel dynamic convolutional block inspired by Vision Outlooker, improving quality and efficiency.

Findings

WOLONet outperforms SOTA vocoders in quality with fewer parameters.

Ablation study confirms the effectiveness of the novel design.

Subjective and objective evaluations show superior synthesis quality.

Abstract

Recently, GAN-based neural vocoders such as Parallel WaveGAN, MelGAN, HiFiGAN, and UnivNet have become popular due to their lightweight and parallel structure, resulting in a real-time synthesized waveform with high fidelity, even on a CPU. HiFiGAN and UnivNet are two SOTA vocoders. Despite their high quality, there is still room for improvement. In this paper, motivated by the structure of Vision Outlooker from computer vision, we adopt a similar idea and propose an effective and lightweight neural vocoder called WOLONet. In this network, we develop a novel lightweight block that uses a location-variable, channel-independent, and depthwise dynamic convolutional kernel with sinusoidally activated dynamic kernel weights. To demonstrate the effectiveness and generalizability of our method, we perform an ablation study to verify our novel design and make a subjective and objective comparison with typical GAN-based vocoders. The results show that our WOLONet achieves the best generation quality while requiring fewer parameters than the two neural SOTA vocoders, HiFiGAN and UnivNet.