How Shrutlekh is Shaping the Future of Multilingual AI in India

India is reshaping AI to reflect how a billion people speak, work, negotiate, govern, and collaborate. In a country where a single conversation can move across Hindi, English, Tamil, Bengali, Marathi, Malayalam, and multiple dialects, AI systems must understand and respond across languages in real time.

Shrutlekh is being developed as part of this effort under the Digital India BHASHINI Division of the Digital India Corporation (DIC), Ministry of Electronics and Information Technology (MeitY). It is designed to function as a foundational language infrastructure layer for governance, enterprise, public communication, and international engagement.

The name ‘Shrutlekh’ comes from the Sanskrit word ‘Shrut’, meaning ‘that which is heard’, and ‘Lekh’, which literally translates to ‘writing what is heard’. In practice, the system converts live speech into structured, searchable, and translatable digital text in real time.

Shrutlekh operates within the broader BHASHINI language AI ecosystem. It integrates Automatic Speech Recognition, language detection, and neural machine translation into a single system. The platform listens to speech, detects shifts in language within the same conversation, generates real-time transcripts, and produces translations without interrupting the flow of speech.

The system is designed to improve documentation and communication workflows. Meetings can generate real-time transcripts and structured reports. Minutes of meetings can be produced during the session itself. Reports and documentation become searchable immediately, allowing teams to review, verify, and circulate records faster.

Its practical applications extend across multiple public and institutional contexts. A citizen who speaks only Bengali attending a consultation conducted in Hindi can follow the proceedings in real time. Officials can understand responses delivered in another language without delay. In courtrooms, witness statements delivered in regional languages can be transcribed and translated instantly, creating digital records that can be archived and reviewed. In healthcare settings, patients describing symptoms in local dialects can be documented more accurately. In classrooms, training sessions, and skilling programs, participants from different states can engage without language becoming a barrier.

Amitabh Nag, CEO, Digital India BHASHINI Division, said, “Building AI for a country as linguistically diverse as ours requires a fundamentally new approach to language technology.”

“Through the BHASHINI ecosystem, we are creating a digital public infrastructure for languages that enables seamless communication, drives innovation across sectors, and ensures that the benefits of AI reach every citizen,” he added. 

Shrutlekh has already been deployed in several public events. During Veer Bal Diwas 2025 celebrations in Puducherry, it translated the Prime Minister’s speech from Hindi to Tamil in real time. At the World Summit on Disaster Management 2025 held in Dehradun, it supported multilingual participation for delegates. It was also used at the 10th TiE Global Summit, Rajasthan Chapter, where it enabled transcription and translation during sessions and supported the retelecast of the PM’s AI Action Summit address from Paris.

The system has also supported over 50 sessions at the India AI Impact Summit and has been used at the RBI Hindi Seminar, the AI-Led Innovation and Capacity Building Conference, the AI4Agri summit 2026, the Republic Day felicitation of 400 Lakhpati Didis, and the Kerala Regional AI Impact Conference 2026. These deployments span governance, finance, agriculture, entrepreneurship, and public policy discussions.

Shrutlekh is also expanding to support more than 35 international languages. This enables multilingual participation in global conferences, research collaborations, and trade engagements. Indian startups can pitch in Hindi while investors follow in English or other languages. Regional entrepreneurs can interact with international partners without relying on interpretation teams.

The system is designed to improve through structured updates. Domain-specific glossaries are used to improve translation accuracy in areas such as law, medicine, finance, agriculture, and governance. These glossaries ensure technical terms are transcribed and translated consistently.

The system also incorporates dialects, accents, and code-mixed speech patterns into its training data. These inputs allow the models to adapt to real conversational usage rather than only formal language structures. Data generated from deployment feeds back into the system, improving recognition accuracy and contextual understanding over time.

Future applications include voice-first government service portals, multilingual conversational AI for citizen services, real-time translation in courts, improved clinical documentation in hospitals, multilingual communication across enterprises operating in different states, and live translation for media broadcasts.

As voice becomes a primary digital interface for millions of users, systems like Shrutlekh provide the underlying infrastructure to ensure that language is not a barrier to participation in governance, commerce, education, or public services.