AI Applications
Bhashini Is Developed By Whom? The Direct Answer, Plus What Most People Miss
Bhashini is built by the Digital India Bhashini Division under the Ministry of Electronics and Information Technology. But the answer is layered: the translation and speech models come from IIT Madras, C-DAC, IIIT Hyderabad, startups and millions of crowdsourced voice samples. Here is who built what, where you have already used Bhashini without knowing, and why its free APIs matter if you want to build Indian-language AI products.
CSEWhy ·
Bhashini is built by the Digital India Bhashini Division under the Ministry of Electronics and Information Technology. But the answer is layered: the translation and speech models come from IIT Madras, C-DAC, IIIT Hyderabad, startups and millions of crowdsourced voice samples. Here is who built what, where you have already used Bhashini without knowing, and why its free APIs matter if you want to build Indian-language AI products.
The direct answer
Bhashini is developed by the Digital India Bhashini Division (DIBD), which sits under the Ministry of Electronics and Information Technology (MeitY), Government of India. It runs as an Independent Business Division of the Digital India Corporation, and it is the delivery arm of the National Language Translation Mission announced in the Union Budget. Prime Minister Narendra Modi launched the platform publicly at Digital India Week in Gandhinagar in July 2022.
So if you are filling an exam answer or a quiz, the expected answer is MeitY. Amitabh Nag has led the division as CEO through most of its build-out. That is the clean answer. The more interesting one is that MeitY did not write most of the AI.
Who actually built the models
Bhashini is better understood as a government-owned platform layer sitting on top of research that was happening anyway. MeitY's job was funding, coordination, the API layer and the mandate to get it into government services. The actual translation, speech recognition and text-to-speech models came from Indian research labs and companies that contribute through ULCA, the Universal Language Contribution API, which works as the open repository and benchmarking system for the ecosystem.
The single biggest technical contributor is AI4Bharat at IIT Madras. Their IndicTrans2 translation models and Indic speech datasets are the backbone of a large part of what Bhashini serves. C-DAC brings decades of Indian language computing work, from fonts and OCR to speech synthesis. IIIT Hyderabad contributes on speech and NLP. And then there is Bhashadaan, the crowdsourcing programme where ordinary people record sentences, type translations and validate other people's contributions, because no Indian language model gets good without data in accents that actual Indians have.
| Contributor | What they bring |
|---|---|
| Digital India Bhashini Division, MeitY | Mission ownership, funding, platform, public APIs, government integrations |
| AI4Bharat, IIT Madras | IndicTrans2 translation models, Indic ASR and speech datasets |
| C-DAC | Long-standing Indian language tech, TTS, OCR, transliteration |
| IIIT Hyderabad and other labs | Speech recognition and NLP research contributions |
| Startups and MSMEs via ULCA | Models and services submitted, benchmarked and served through the platform |
| Citizens via Bhashadaan | Crowdsourced voice, text and validation data across languages |
A short timeline
The mission did not appear overnight, and the dates explain why the platform looks the way it does. It was conceived as public digital infrastructure, in the same family of thinking as UPI and Aadhaar authentication: one government-run layer, open APIs, everyone else builds on top.
| When | What happened |
|---|---|
| Budget 2021-22 | National Language Translation Mission announced as a national priority |
| July 2022 | Bhashini launched publicly at Digital India Week, Gandhinagar |
| 2023 | Bhashini mobile app released; live Hindi-to-Tamil translation of the PM's speech at Kashi Tamil Sangamam |
| 2024 onwards | Integration into government apps, helplines, railway services, courts and state portals |
You have probably already used it
Most people who search for who developed Bhashini have no idea they have touched it. If you have used a government app that read out a form in Marathi, or a helpline that understood a spoken query in Bhojpuri, or a court or railway service that produced a translated document, there is a reasonable chance Bhashini's APIs were doing the work behind the screen.
That is the real design intent. Bhashini was never meant to be a consumer product competing with Google Translate on polish. It was meant to be plumbing. A state government department, a bank, a health scheme or a startup plugs into the API and suddenly supports 22 scheduled languages without hiring a single NLP engineer.
It is not uniformly good. Hindi, Tamil, Bengali and Marathi outputs are usable. Lower-resource languages are noticeably weaker, and anyone who has tested Bhashini on technical or legal text knows it still needs a human reviewer. That gap is the point of Bhashadaan.
Why this matters if you want to build something
Here is the part worth your attention. Bhashini's APIs are open and free to use for a very large range of cases, which means the hardest and most expensive piece of building an Indian-language product has already been paid for by the government. Translation, speech-to-text, text-to-speech, transliteration. You do not have to train any of it.
Think about what that enables. A WhatsApp bot for a farmer producer organisation that takes voice notes in Marathi and replies in Marathi. A coaching institute that turns one English lecture into subtitled versions in four languages. A local insurance agent whose claim-form explainer runs in Telugu audio. These are not research projects. They are weekend builds for someone who knows how to call an API and wire it into a workflow, and they are exactly the kind of thing Indian businesses will pay for, because the alternative is hiring people who speak the language.
The gap is almost never the technology. It is that very few people have actually built an end-to-end thing that calls a model, handles messy real-world input and runs reliably for someone else. If you are a student or early-career professional who wants to cross that gap by shipping projects instead of collecting course certificates, the AI Creator Fellowship is built around that: eight weeks, hands-on building, automation workflows and a portfolio you can show. Businesses sitting on repetitive multilingual work, like support queues or document processing, usually need the same capability delivered as a custom automation rather than a course.
The government built the language layer. Nobody is coming to build the applications on top of it for you.
FAQs
1. Is Bhashini free to use?
Yes, Bhashini's translation, speech-to-text and text-to-speech APIs are available free for most use cases after you register on the platform and generate API keys. Usage limits and terms apply, and high-volume commercial deployments should check the current policy on the official portal.
2. How many languages does Bhashini support?
Bhashini targets all 22 scheduled languages of India, though coverage quality varies sharply by language and by task. Hindi, Bengali, Tamil, Telugu, Marathi and Kannada are far better served than low-resource languages, where models are still improving through crowdsourced data.
3. Is Bhashini the same as Google Translate?
No. Google Translate is a consumer product run by a private company, while Bhashini is Indian government-owned public infrastructure whose main purpose is to let other apps, departments and businesses plug language capability into their own services through APIs.
4. Who is the CEO of Bhashini?
Amitabh Nag has served as the CEO of the Digital India Bhashini Division, the MeitY body that runs the platform. The division operates as an Independent Business Division under the Digital India Corporation.
5. What is Bhashadaan?
Bhashadaan is Bhashini's crowdsourcing initiative where volunteers contribute voice recordings, translations, transliterations and image labels in Indian languages, and validate contributions from others. The data collected is used to improve the accuracy of Bhashini's models, especially for languages with little digital text available.
6. Can developers build commercial products using Bhashini APIs?
Yes, developers and startups can integrate Bhashini APIs into their applications, and many Indian products already use it for multilingual voice and text features. Register on the Bhashini developer portal, get your credentials, and check the current terms for commercial and high-volume usage.