toiar/Rynsan-TTS

Rynsan TTS

Rynsan TTS is a multilingual text-to-speech model that has been extended to support Khasi, Garo, and Pnar, languages of Meghalaya that have historically had limited representation in modern speech technology.

The broader goal of Rynsan is to build accessible speech technology for the diverse languages and dialects of Meghalaya, with a particular focus on Khasic languages, supporting their preservation, development, and use in modern voice-based applications.

Developed under the Tynrai AI initiative.

🎙️ Try Rynsan TTS: Live Demo

Model Details

Attribute	Details
Model	Rynsan TTS
Repository	toiar/Rynsan-TTS
Base Model	k2-fsa/OmniVoice
Developer	Toiar / Tynrai AI
Task	Text-to-Speech
Model Type	Multilingual TTS
Languages	Multilingual, including English, Hindi, Khasi, Garo, and Pnar

Language Extension

Rynsan TTS builds upon the multilingual capabilities of k2-fsa/OmniVoice and extends the model to better support low-resource languages of Meghalaya.

The current Rynsan extension focuses on:

Language	ISO 639-3
Khasi	kha
Garo	grt
Pnar	pbv

The underlying OmniVoice model supports a broader range of languages. Rynsan retains the multilingual capabilities of the base model while extending its capabilities to languages that have substantially fewer speech resources.

Uses

Intended Use

Rynsan TTS is intended for:

• Multilingual text-to-speech applications
• Text-to-speech applications in Khasi, Garo, Pnar, English, Hindi, and other supported languages
• Accessibility and assistive technologies
• Educational and language-learning tools
• Digital content creation
• Voice interfaces and conversational applications
• Research on multilingual and low-resource TTS
• Research on speech technology for languages and dialects of Meghalaya
• Preservation and digitization of underrepresented languages
• Development of language-specific speech datasets and resources

Training and Model Development

Rynsan TTS is an extension of the multilingual k2-fsa/OmniVoice model.

Rather than building a multilingual TTS system from scratch, Rynsan builds upon the capabilities of the existing multilingual foundation and extends them with additional speech data for Khasi, Garo, and Pnar.

This approach aims to make multilingual speech technology more useful for languages of Meghalaya that have comparatively limited speech resources.

The broader goal of Rynsan is to progressively support the diverse languages, dialects, and speech varieties of Meghalaya, with particular attention to Khasic languages and other underrepresented language communities.

Limitations

Potential limitations include:

• Pronunciation errors for uncommon or unfamiliar words
• Inconsistent pronunciation of names, loanwords, and newly introduced terminology
• Code-switching between languages may produce inconsistent results
• Regional and dialectal pronunciation differences may not be fully represented
• Some dialects or speech varieties may have limited representation in the training data
• Numbers, abbreviations, symbols, and special characters may require text normalization
• Speech quality may vary depending on the input text and language
• Performance on the extended low-resource languages may differ from languages with substantially larger speech resources

Rynsan should not be assumed to represent every dialect, accent, or speaking style within a language.

Citation

@misc{rynsan_tts,
  title        = {Rynsan TTS},
  author       = {Toiar},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {https://huggingface.co/toiar/Rynsan-TTS}
}