Navana.ai launches Bodhi TTS, a text-to-speech model for Indian languages
What's the story
Navana.ai has launched Bodhi TTS, a revolutionary text-to-speech (TTS) voice AI model. The innovative technology is designed for real-time applications in 10 languages including Hindi, Telugu, Tamil, Marathi, Bengali, Odia, Malayalam, Kannada, English, and Gujarati. The company has priced the service at ₹12 per 10,000 characters of generated speech, or around ₹0.72 per minute of generated speech.
Features
The service can generate audio in under 100 milliseconds
Bodhi TTS can generate audio in less than 100 milliseconds.
It is capable of handling Indian business-specific text such as currency values, PAN numbers, phone numbers, dates, EMIs, scheme names, and account numbers.
The model also comes with pronunciation controls that let enterprises upload dictionaries with product names, branch names, and customer names. These changes can be made without retraining the model.
Advanced capabilities
It also supports zero-shot voice cloning
Bodhi TTS also supports zero-shot voice cloning. This means users can upload a few seconds of reference audio via the API to create a custom voice without additional training.
The company claims that reference audio is cached so custom voices don't add latency during calls.
Navana.ai says Bodhi TTS has been built and trained in-house, making it five times smaller than other comparable architectures.
Security measures
The model can run within an enterprise's data center
The compact size of Bodhi TTS allows it to run within an enterprise's own data center, keeping customer audio, transcripts, and personal data within its security environment.
The company also offers new users ₹1,000 in free credits on signup for their platform.
Banks, insurers, and lenders can apply for early access to Bodhi TTS and Bodhi STT. The first 100 approved teams will get 100 hours of usage credits valid for 90 days.