India live-translates Modi's speech into 22 languages by AI

For the first time at national scale, India ran Prime Minister Narendra Modi's Independence Day speech on August 15 through an AI translation pipeline: the government's open-source platform Bhashini rendered the Red Fort address simultaneously into 22 Indian languages. TechTimes calls it the "first Independence Day address processed through an AI translation pipeline at national scale".

Technically, Bhashini works in three stages: speech recognition converts the Hindi original into text, the neural translation system IndicTrans2 transfers it into the target languages, and text-to-speech turns that into audio. Training data includes around 230 million bilingual text pairs from the Bharat Parallel Corpus Collection covering all 22 scheduled languages.

Politically, Modi tied the appearance to ambitious technology goals: 10 million young Indians are to receive AI skills training over the coming year. For context: by the end of June, the main government program had certified only about 25,000 candidates in AI roles — the pledge means multiplying the current pace many times over.

Added to that are five to eight planned semiconductor plants within seven to eight years and a target of 100 gigawatts of nuclear capacity by 2047. India is positioning itself as an AI nation building its own stack — from language models to chips — rather than merely importing.

For the Global South, Bhashini is a reference case: a government using open AI infrastructure to turn multilingualism from a barrier into a feature. Expect imitators — and a wider market for multilingual AI services.