A technological sleight of hand—convincing enough to fool millions
In the age of artificial intelligence, even the voices of the dead can be summoned to deceive the living. A viral video circulating in India falsely claimed that the hit song from the 2025 film Saiyaara had been stolen from a lost Kishore Kumar recording — a legend whose voice shaped generations of Indian cinema. Fact-checkers traced the audio not to any archive, but to a content creator who had used AI to clothe a radio jockey's vocals in Kumar's unmistakable style, pairing the fabrication with unrelated vintage footage to manufacture the illusion of history.
- Thousands of social media users shared the video convinced they had uncovered a case of cultural theft — that a beloved classic had been quietly plundered by a modern Bollywood production.
- The footage of Amitabh Bachchan and Moushumi Chatterjee lent the claim a visual authority it did not deserve, borrowed from a 1979 film entirely unrelated to the Saiyaara soundtrack.
- Fact-checkers dismantled the claim in two steps: reverse image search exposed the footage as a scene from Manzil's 'Rimjhim Gire Sawan,' and an Instagram post revealed the AI-generated audio's true origin.
- Content creator Anshuman Sharma openly disclosed in his own captions that he had used artificial intelligence to transform RJ Kisna's vocals into a convincing imitation of Kishore Kumar's voice and style.
- No recording of Kishore Kumar singing Saiyaara exists anywhere in his discography — the entire viral narrative was built on a technically sophisticated but entirely fabricated foundation.
In late July, a video began spreading across Indian social media showing what appeared to be a vintage Bollywood scene — Amitabh Bachchan and Moushumi Chatterjee on screen, a song playing beneath that sounded unmistakably like Kishore Kumar. The claim attached to it was incendiary: the recently released film Saiyaara had stolen this song, and the original belonged to the legendary playback singer.
Fact-checkers began with the visuals. A reverse image search traced the footage to Shemaroo Entertainment's official YouTube channel — not to any Saiyaara-related material, but to 'Rimjhim Gire Sawan' from the 1979 film Manzil. The clip had simply been repurposed to create a false impression of origin.
The audio trail led somewhere equally revealing. An Instagram post from July 26 by content creator Anshuman Sharma showed the Saiyaara song rendered in what sounded like Kishore Kumar's voice — but Sharma's own caption explained exactly what it was: an AI-generated fabrication. He had taken vocals originally sung by radio jockey RJ Kisna and used artificial intelligence to recast them in Kumar's tonal style, then layered the result over newly produced retro-styled instrumentation.
Sharma confirmed the process to PTI Fact Check in full. No original Kishore Kumar recording of Saiyaara exists. What millions heard was a technological imitation — convincing enough to pass as lost cinema history, but constructed entirely in the present. The episode stands as a sharp illustration of how AI-generated audio, paired with authentic-looking visuals, can outpace truth in the time it takes a video to go viral.
A video began circulating on social media in late July showing what appeared to be a scene from an old Bollywood film—Amitabh Bachchan and Moushumi Chatterjee on screen, a song playing softly underneath. The song was "Saiyaara Tu Toh Badla Nahin Hai, Mausam Zara Sa Rootha Hua Hai," and it sounded like it was sung by Kishore Kumar, the legendary playback singer whose voice defined an era of Indian cinema. Users posting the clip made a bold claim: this was the original version, they said, and the recently released Bollywood film Saiyaara had stolen it. The song, they insisted, belonged to Kumar, not to the new movie.
When fact-checkers began investigating, they started with the visuals. Using reverse image search on key frames from the viral video, they traced the footage to an official YouTube channel run by Shemaroo Entertainment Limited. The images matched perfectly—same actors, same scene. But the YouTube description told a different story. The clip was not from some forgotten Kishore Kumar recording. It came from "Rimjhim Gire Sawan," a song from the 1979 film Manzil. The video had nothing to do with Saiyaara at all.
That discovery opened a second line of inquiry. Searching for connections between Kishore Kumar and the Saiyaara track, investigators found a post dated July 26 on Instagram from a user named Anshuman Sharma. There it was: the Saiyaara song rendered in what sounded like Kishore Kumar's voice. But in the caption, Sharma was explicit about what he had done. The vocals, he wrote, had been created using artificial intelligence. This was not a lost recording. It was a digital fabrication, skillfully made but fabricated nonetheless.
A longer version of the same track appeared on Sharma's YouTube channel, and again the description made clear what had happened. "The full version is finally here!" Sharma wrote. "We converted the vocals sung amazingly by RJ Kisna into Kishore da's voice using AI, and I produced the retro background music." The original singer was RJ Kisna, a radio jockey. Sharma had taken those vocals and run them through artificial intelligence software designed to mimic the tonal qualities and style of Kishore Kumar. He had also produced new instrumental backing, styled to sound like a retro recording from decades past.
When contacted by PTI Fact Check, Sharma confirmed the entire process. RJ Kisna had sung the original vocals for the Saiyaara track. Sharma then used AI technology to transform those vocals into something that resembled Kishore Kumar's voice and singing style. The background music was newly produced by Sharma himself, designed to evoke the sound of older film songs. No original Kishore Kumar recording of Saiyaara exists anywhere in his discography. The song is entirely new, and what millions of people heard in that viral video was a technological sleight of hand—convincing enough to fool social media users into believing they had discovered a lost piece of cinema history.
The fact-check revealed a broader pattern: the video clip itself, showing Bachchan and Chatterjee, had no connection to the Saiyaara soundtrack at all. It was simply old footage paired with new audio, creating the illusion of authenticity. In an era when AI-generated content can sound nearly indistinguishable from the real thing, the viral claim had spread quickly, each share reinforcing the false narrative. By the time fact-checkers caught up, thousands of people had already accepted the story as true. The investigation stands as a reminder that viral audio, no matter how convincing, demands verification before it becomes accepted as history.
Notable Quotes
We converted the vocals sung amazingly by RJ Kisna into Kishore da's voice using AI, and I produced the retro background music.— Anshuman Sharma, content creator
The Hearth Conversation Another angle on the story
How did something this convincing spread so widely before anyone checked it?
The video had two things working in its favor: it featured recognizable actors from a real old film, and the AI voice was genuinely well-made. People saw Bachchan and Chatterjee and assumed the audio matched. They didn't question whether the footage and the song actually belonged together.
But someone had to know this was AI-generated, right? Why post it without that context?
Anshuman Sharma did label it clearly on his own channels—he was transparent about using AI. But once the clip got extracted and reshared, that context disappeared. It became just a video with no attribution, no explanation. By then it was already a story.
What made the AI voice so convincing that people believed it was Kishore Kumar himself?
The technology has gotten very good at capturing not just the pitch and timbre of a voice, but the phrasing, the emotional weight. Kishore Kumar had such a distinctive style that if you get those elements right, people's ears believe what they're hearing. Add in the nostalgia of old film footage, and the brain fills in the rest.
Does this change how we should think about old recordings or archival material online?
It should. If something sounds too perfect, too timely, too good to be true—especially if it's audio—it's worth asking where it came from. A YouTube description or an Instagram caption matters now in a way it didn't before. The metadata is part of the story.
What happens to the people who shared this before the fact-check came out?
Most probably never see the correction. That's the real problem. The false version travels fast and far. The truth travels slower and reaches fewer people. The viral claim becomes the accepted version in people's minds.