Shunya Labs, in partnership with Nasscom, Launches Vāķ, India’s Largest Open-Weight Voice AI ......
Shunya Labs, a Nasscom GenAI Cohort 1 startup, has announced the launch
of Vāķ, India’s largest open-weight voice AI release, unveiled in partnership
with Nasscom at the India AI Impact Summit 2026 at Bharat Mandapam, New Delhi.
A real-time translation system spanning all 2,970 language pairs across India's
55 mother tongues, Vāķ is accompanied by the world's #1-ranked open-weight
speech recognition model and a neural text-to-speech engine, all released
simultaneously, all free to deploy.
In a
country of 1.4 billion people with over a thousand language varieties,
including 61 languages spoken by more than a million people, speech AI systems
that support only 5–10 Indian languages are simply inadequate. This limited
coverage has compelled Indian developers, start-ups, and government
institutions to rely on foreign APIs and overseas servers, placing citizen
voice data under external terms while leaving vast linguistic communities
underserved. Additionally, many Indian languages are not mutually intelligible,
even within the same language family, making genuine linguistic inclusion far
more complex than superficial multi-language support. As a result, major speech
communities such as Bhojpuri, Rajasthani, Chhattisgarhi, and Magahi each spoken
by tens of millions have remained largely excluded from mainstream AI systems.
Vāķ is designed to close this structural gap by delivering comprehensive,
real-time, and sovereign language coverage aligned with the full depth and
diversity of India’s linguistic landscape.
At its core, Vāķ is a real-time Any-to-Any Translation system that
converts speech in any of India's 55 languages into any of the other 54,
instantly, with end-to-end latency under 1.5 seconds. But it does more than
merely translate words; it preserves the speaker’s voice, their tone, and their
emotions across languages. Additionally, its Zero-shot voice cloning means that
no prior training data is needed.
The name Vāķ itself carries deep resonance, representing the power of
speech in ancient Vedic thought. It’s the force that transforms thought into
expression and gives voice to knowledge itself, and one of the oldest and most
revered concepts in Sanskrit tradition. Shunya Labs’ choice of the name
reflects how Vāķ is a technology that is rooted in India’s own intellectual
heritage. With complete model weights publicly available for download and local
deployment, any organisation can run Vāķ on its own infrastructure with no data
leaving its environment.
The release comprises three simultaneously launched model families.
Pingala ASR holds the #1 rank on the Hugging Face OpenASR Leaderboard, a
globally recognised, third-party benchmark achieving a Word Error Rate of
3.10%, the lowest ever recorded, with a CPU-first architecture delivering
sub-250ms latency for edge and offline deployment. Vāķ’s neutral TTS delivers
natural-sounding output across all 55 languages with streaming capability,
custom voice creation, zero-shot voice cloning, and prosody and emotion
control. Real-Time Any-to-Any Translation enables translation across all 2,970
language pairs (55 × 54), preserving the speaker's voice and emotional
register, and end-to-end latency under 1.5 seconds.
It’s these first-of-its-kind features that give Vāķ its edge. Vāķ gives
any organisation the freedom to deploy entirely on their own infrastructure
without calls to foreign servers, or citizen voice data leaving the country, no
recurring API costs, and no dependency. For government agencies, hospitals
reaching rural patients, and courts delivering justice in local languages,
sovereign deployment is the foundation. With Vāķ, the full set of 2,970
translation pairs is available out of the box, making cross-linguistic citizen
services, healthcare outreach, judicial access, and educational delivery
technically and economically viable for the first time at a national scale.
Commenting on the launch, Sourav Bandyopadhyay, Founder &
Chief Scientist, Shunya Labs, said, “India does not need to depend on
foreign APIs to hear its own people. With Vāķ, we are releasing the world's
best-performing open-weight Voice AI across 55 distinct Indian tongues, that
is, every language with over a million speakers. This is sovereign, self-funded
innovation built within the Nasscom ecosystem. Our vision is simple: every
developer can build, every government can deploy, and every Indian can be heard
in their own language.”
Adding to this, Ankit Bose, Head of AI, Nasscom, said,
“Shunya Labs, as a Nasscom GenAI Cohort 1 startup, exemplifies the kind of
globally competitive innovation emerging from India's AI ecosystem. Building
high-performance, open-weight models across Indian languages reflects both
technical excellence and a commitment to sovereign, inclusive AI. The
introduction of Vāķ reinforces the role of the Nasscom GenAI ecosystem in
enabling startups to innovate at global standards while contributing
meaningfully to India's digital future.”
Vāķ’s coverage spans India's full linguistic landscape: 43 Indo-Aryan
languages including Hindi, Bengali, Marathi, Gujarati, Urdu, Bhojpuri, and
Rajasthani; seven Dravidian languages including Telugu, Tamil, Kannada, and
Malayalam; three Sino-Tibetan languages including Meitei (Manipuri), Bodo, and
Garo; one Austroasiatic language, Santali; and Indian English, reaching over
1.17 billion native speakers in aggregate. The launch of Vāķ also aligns with
the IndiaAI Mission, Digital India, BHASHINI, and Atmanirbhar Bharat.
Leave A Comment