Skip to content
AIStartup NewsNews

Smallest.ai Launches AWAAZ to Address Quality and Cost Issues in Indian TTS Models

AWAAZ boasts one of the Mean Opinion Score (MOS) in Hindi and Indian English, supports over 10 accents, features single-shot voice cloning using just a 5-second audio clip, and also offers a 200ms streaming latency.

  • Sudarshan Kamath, founder of smallest.ai, announced the beta launch of AWAAZ, a multilingual, multi-accent text-to-speech model for Indian languages.
  • AWAAZ, priced at Rs. 999 for 500,000 characters, aims to solve quality and cost issues in text-to-speech models for Indian languages.

Co-founders of smallest.ai, Sudarshan Kamath and Akshat Mandloi recently announced today on LinkedIn the beta launch of AWAAZ, a multilingual, multi-accent text-to-speech model designed for Indian languages.

AWAAZ boasts a state-of-the-art Mean Opinion Score (MOS) in Hindi and Indian English, supports over 10 accents, features single-shot voice cloning using just a 5-second audio clip, and also offers a 200ms streaming latency. 

According to Kamath, Smallest.ai created AWAAZ in response to the challenges faced when searching for text-to-speech (TTS) models for building voice bots in Indian languages. The existing models were either of subpar quality or prohibitively expensive for scaling up. This issue arises because the quality of speech data for TTS systems decreases significantly when moving from English to other languages. 

“When we started looking for TTS models for building voice-bots for Indian languages, we realized that the current models are either: a) subpar in terms of quality or b) too expensive for scaling up,” wrote Kamath.

Founded in 2023, smallest.ai's AWAAZ aims to address this by utilizing high-quality multi-language, multi-accent datasets, and establishing a platform to continue developing TTS solutions for additional languages, with a focus on India and South Asia. 

Moreover, as an introductory offer, AWAAZ will be priced at Rs. 999 for 500,000 characters, significantly lower than its competitors.


Edited by Harshajit Sarmah

Latest