How is Shunya Labs' New Translation System Impacting 55 Indian Languages?
Synopsis
At the India AI Impact Summit, Shunya Labs unveiled Vaķ, a groundbreaking voice AI and translation system that supports 55 Indian languages. This innovative technology promises to revolutionize communication and access to services, making it a game-changer in the realm of language translation.
Key Takeaways
Supports 55 Indian languages Real-time translation capabilities Zero-shot voice cloning technology Sub-250ms latency for edge and offline deployment Empowers government and healthcare services
New Delhi, Feb 19 (NationPress) During the India AI Impact Summit, Shunya Labs, a startup from the Nasscom GenAI cohort, officially unveiled Vaķ, an advanced open-weight voice AI system that facilitates real-time translations across 55 Indian languages and 2,970 language pairs.
According to the company's announcement, Vaķ integrates the leading open-weight speech recognition model and a neural text-to-speech engine, all available for free deployment.
The system is designed to provide real-time “Any-to-Any Translation”, allowing users to convert speech from any of the 55 supported languages into any of the other 54, achieving an impressive end-to-end latency of under 1.5 seconds.
Shunya Labs highlighted that Vaķ employs zero-shot voice cloning, eliminating the need for prior training data. Its neutral TTS offers natural-sounding audio in all 55 languages, featuring streaming capabilities, custom voice generation, and emotional control, as mentioned in the statement.
The platform effectively preserves the original speaker's voice and emotional tone.
Notably, the Pingala ASR has been recognized as the top model on the Hugging Face OpenASR leaderboard, achieving an unprecedented 3.10 percent word-error rate.
Its CPU-first architecture ensures sub-250ms latency for edge and offline usage. The complete model weights are publicly accessible for local deployment, allowing organizations to operate the models on-premises without transmitting voice data to external servers.
This capability supports sovereign deployment for government entities, healthcare providers serving rural populations, and courts facilitating justice in local languages.
According to the release, Vaķ has made cross-lingual citizen services, healthcare access, judicial support, and educational delivery feasible on a nationwide scale.
“This is a self-funded, sovereign innovation developed within the Nasscom ecosystem. Our vision is clear: every developer can create, every government can implement, and every Indian can communicate in their preferred language,” stated Sourav Bandyopadhyay, Founder & Chief Scientist of Shunya Labs.
Vaķ encompasses 43 Indo-Aryan languages such as Hindi, Bengali, Marathi, Gujarati, Urdu, Bhojpuri, and Rajasthani, along with 7 Dravidian languages, 3 Sino-Tibetan languages, and 1 Austroasiatic language, plus Indian English, collectively catering to over 1.17 billion native speakers across the country.
aar/pk
Point of View
The launch of Vaķ by Shunya Labs signifies a noteworthy advancement in India's technology landscape. This tool not only enhances communication among diverse linguistic groups but also empowers government services and healthcare in rural areas. As we embrace such innovations, it is essential to ensure that technology is developed and deployed in a manner that is accessible and beneficial for all citizens.
NationPress
10 May 2026
Frequently Asked Questions
What technologies power Vaķ?
Vaķ is powered by the leading open-weight speech recognition model, a neural text-to-speech engine, and features zero-shot voice cloning.
Who can benefit from Vaķ?
Vaķ is designed for a wide range of users, including government agencies, healthcare providers, and educational institutions, facilitating communication in local languages.
What makes Vaķ unique?
Vaķ's unique features include its ability to preserve the speaker's voice and emotional tone, as well as its publicly accessible model weights for local deployment.
