Bhuvaneswari’s Audio: Forensic reports rule out deepfake; source still unknown

Hyderabad: An audio of TDP chief Chandrababu Naidu’s wife Nara Bhuvaneswari allegedly using offensive remarks is viral on social media. An X premium handle, YS Jagan Trends, shared the video claiming that Bhuvaneswari’s derogatory remarks were directed towards Dalits.

The viral audio led to chatter on social media with people condemning the alleged insult by Bhuvaneswari to the Dalit community. The official handle of the Telugu Desam Party issued a clarification stating that the audio is a deepfake. The handle accused Andhra Pradesh chief minister YS Jagan Mohan Reddy of creating the fake audio.

Nara Lokesh, national general secretary of TDP and son of Bhuvaneswari, accused CM Jagan of fabricating the fake audio. He further attacked CM Jagan claiming that a man who could take a dig at his own sister for her attire can also stoop low to defame Lokesh’s mother.

What does the audio contain?

In the 57 seconds-long audio, a woman's voice could be heard. She is audibly angry with one or more persons whose voice(s) is not heard in the audio file. The woman could be heard using expletives in Telugu and English, including calling them as being ‘born in a dustbin’ while she is ‘born in a house bigger than them.’ There is no mention of the word ‘Dalit’.

At the end of the video, the voice of the woman speaker echoes as if the person is speaking inside a hall. The drastic change in sound acoustics may indicate that the audio might have been tinkered.

Audio analysed by industry experts

The Misinformation Combat Alliance (MCA), an Indian cross-industry collaborative effort to combat the spread of misinformation that includes NewsMeter as one of its members, subjected the audio to analysis through its Deepfakes Analysis Unit (DAU).

The Deepfakes Analysis Unit (DAU) forwarded the audio to its expert partners, including Dr Hany Farid’s forensic lab and ContrailsAI. Their analysis confirmed that the audio was not generated using Artificial Intelligence (AI).

As per Berkeley’s website, Dr Hany Farid is a professor at the University of California, Berkeley with a focus on digital forensics, forensic science, misinformation, image analysis, and human perception.

Farid’s lab confirmed that the accent and intonation in the audio did not seem odd. They added that an AI-generated audio does not convincingly imitate strong emotions like anger and aggression. The lab, however, did not confirm that the audio could be of Nara Bhuvaneswari.

ContrailsAI, a Bangalore-based startup that has developed its own multimodal A.I. tools for detection of images,audio and video deepfakes also examined the audio by creating audio clones for various Telugu politicians. They found audio clones could not have the kind of voice quality and intonations that could be heard in the alleged Bhuvaneswari’s audio. Additionally, they conducted a subjective experiment on the audio, leading to the conclusion that it was mostly real not a deepfake.

No confirmation of identity of the speaker or source of the audio

However, the laboratory and the tech startup did not confirm that the viral video could be of Nara Bhuvaneswari. It’s possible that this is a case of digital fabrication, where an audio of another woman has been falsely attributed to Bhuvaneswari. NewsMeter also couldn’t independently verify the source of the audio and the identity of the speaker.