E-Paper
AI model's shayari fools millions of viewers, video gets 28 million views. It was copied from…
A viral AI-generated video, featuring a woman reciting a couplet, managed to fool millions of people.
A woman in a yellow kurta takes the microphone onstage, smiles for the audience, and then delivers a couplet in Urdu. “Khaandani raees rakhte hain mijaaz naram apna, khaandani raees rakhte hain mijaaz naram apna… tumhara lehja bata raha hai ki tumhari daulat nayi nayi hai,” she says, ending to thunderous applause.
The piece of shayari roughly translates to: “People with old money act with grace. Your tone makes it clear that your money is brand new.”
The video features and was shared on Instagram by one Tanvi Joshi, whose profile describes her as a “Punjabi girl”. It has gone massively viral online, racking up 28 million views in just one day.
Millions of people came across the shayari video on their Instagram. They commended the wisdom of Joshi’s words and debated about the correct pronunciation of mijaz — “it’s mizaj,” read several comments under the post.
But although millions of people watched and ‘liked’ the video, few realised that they were looking at an AI model.
That’s right — Tanvi Joshi does not actually exist. The AI model has managed to trick millions of viewers with her surprisingly life-like videos, never mind her suspiciously blemish-free skin and the fact that her voice changes from one video to the next.
The real shayari video
The video of Tanvi Joshi reciting a couplet on old vs new money is not original either. The audio was lifted from Marziya Shanu Pathan, a corporator of Thane Municipal Corporation.
Six days ago, Pathan had recited the same couplet at an event and posted the video on Instagram. Ironically, the original video got less than 9 lakh views, compared to Tanvi Joshi’s 28 million views.
Pathan’s voice was used in an AI video that shows ‘Tanvi Joshi’ reciting the same shayari.
“Aye thats my voiceeeeeee,” Pathan posted under Joshi’s viral video. Her comment led many to realise they were looking at an AI video.
The episode underlines just how difficult AI-generated videos are becoming to spot . As synthetic faces, voices and performances grow more convincing, they raise serious questions about consent, attribution and trust: whose words are being used, who gets the credit and how easily real people can be impersonated.
