| Property | Value |
|---|---|
| Author | ai4bharat |
| Model Type | Text-to-Speech (TTS) |
| Languages Supported | 11 Indian Languages |
| Training Data | 1417 hours of speech |
| Model URL | Hugging Face |
What is IndicF5?
IndicF5 is a state-of-the-art polyglot Text-to-Speech model specifically designed for Indian languages. Developed by ai4bharat, it provides near-human quality speech synthesis across 11 major Indian languages including Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu. The model leverages high-quality speech data from multiple datasets including Rasa, IndicTTS, LIMMITS, and IndicVoices-R.
Implementation Details
The model implementation requires Python 3.10 and can be easily installed through pip. It utilizes the Transformers library and requires three key inputs for speech generation: the text to synthesize, a reference prompt audio for voice characteristics, and the transcript of the reference audio.
- Built on the F5-TTS architecture with specific optimizations for Indian languages
- Supports voice cloning capabilities through reference audio
- 24kHz sampling rate output
- Implemented using the Transformers library ecosystem
Core Capabilities
- Multi-lingual speech synthesis across 11 Indian languages
- Near-human quality voice generation
- Voice cloning through reference audio prompts
- High-fidelity audio output with proper prosody handling
- Seamless integration with Python environments
Frequently Asked Questions
Q: What makes this model unique?
IndicF5 stands out for its comprehensive coverage of Indian languages and near-human speech quality. It's specifically optimized for Indian language phonetics and prosody, trained on an extensive dataset of 1417 hours of high-quality speech.
Q: What are the recommended use cases?
The model is ideal for applications requiring high-quality Indian language speech synthesis, including educational content, accessibility tools, voice assistants, and content localization. However, it's important to note that voice cloning requires explicit permission from the voice owner.