Indic Parler-TTS
| Property | Value |
|---|---|
| Developer | AI4Bharat & HuggingFace |
| Languages Supported | 21 languages |
| License | Apache 2.0 |
| Training Data | 1,806 hours multilingual dataset |
What is indic-parler-tts?
Indic Parler-TTS is a groundbreaking multilingual text-to-speech model specifically designed for Indian languages. It's an extension of Parler-TTS Mini, fine-tuned on a comprehensive dataset of 1,806 hours covering 21 languages including major Indian languages and English. The model stands out for its ability to generate natural, high-quality speech with controllable characteristics across multiple languages.
Implementation Details
The model utilizes two distinct tokenizers - one for the prompt and another for the description. It processes text input along with detailed voice descriptions to generate speech with specific characteristics. The system supports 69 unique voices across different languages, with each voice capable of producing variations in pitch, speed, expressiveness, and audio quality.
- Advanced prompt tokenization system with byte fallback capability
- Dual tokenizer architecture for enhanced multilingual support
- Automatic language detection and adaptation
- Support for emotion-specific prompts in 10 languages
Core Capabilities
- Multilingual Support: Official support for 21 languages with high-quality synthesis
- Voice Customization: Control over background noise, reverberation, expressivity, pitch, and speaking rate
- Speaker Diversity: 69 unique voices with recommended speakers for each language
- Emotion Rendering: Support for various emotional tones including command, anger, happiness, and more
- High Performance: Native Speaker Scores ranging from 75% to 99% across languages
Frequently Asked Questions
Q: What makes this model unique?
The model's ability to handle 21 different languages with high native speaker scores, combined with its extensive voice customization options and emotion rendering capabilities, makes it particularly valuable for Indian language technology applications.
Q: What are the recommended use cases?
The model is ideal for applications requiring multilingual text-to-speech conversion, including educational content, accessibility tools, automated customer service, and content localization for Indian languages. It's particularly effective for scenarios requiring natural-sounding speech with specific voice characteristics.
