About this document
Nepali Speech To Text Tacotron by Shreekar is a document available to read on EtoBox.
The paper presents a method for generating high-quality Nepali speech using the Tacotron2 model for melspectrogram generation, addressing limitations in existing TTS systems. It details a two-phase process involving melspectrogram generation and vocoder output, with a new dataset created by the authors to enhance performance. The system achieved a Mean Opinion Score of 4.03 for naturalness, the highest among previous Nepali TTS tasks, demonstrating its effectiveness and contribution to the field.
- Author
- Shreekar
- Language
- EN