The Deepfakes Analysis Unit (DAU) analysed a video that apparently shows Rajat Sharma, India TV news anchor and chairperson, reporting the death of cardiologist Dr. Bimal Chhajer in a supposed car blast. The doctor, however, issued a statement clarifying that he is alive and that synthetic voice was used in his death hoax video to make it sound like his voice. After putting the video through A.I. detection tools and getting experts to weigh in we were able to conclude that synthetic audio was used to fabricate the video.
The three-minute-and-seven-second video in Hindi, embedded in a Facebook post, was escalated to the DAU by a fact-checking partner for analysis. The video is not accessible through that link anymore, however, we do not know if it is still visible through other links on the same platform or elsewhere.
The video opens with Mr. Sharma speaking to the camera in a studio-like setting with a colourful backdrop, Dr. Chhajer’s picture is visible as an inset in the top-left corner of the video frame. A male voice recorded over these visuals announces the death of the doctor alleging the involvement of pharmaceutical companies, following which the video briefly cuts to visuals of a car in flames.
As the narration of the alleged incident continues through the male voice whose supposed face is Sharma, a short clip apparently of a male police officer plays out confirming the doctor’s death in a blast. However, the officer’s face and neck are blurred out. His khaki uniform, blue beret, epaulette, and a microphone bearing a logo resembling that of Asian News International (ANI) — a news agency — are visible in the mid-shot in which the interview has been recorded.
The video ends with a clip featuring an interview of Chhajer, which the male voice accompanying Sharma’s visuals introduces as the final recorded interview of the doctor. In that clip the doctor appears to be speaking into a microphone in an interview-like set-up with only him visible in the frame.
The male voice with the doctor’s visuals claims to have received threats from pharmaceutical companies for coming up with affordable medication to cure cardiovascular ailments in record time. That voice urges viewers to get details on the product by pressing a button, which is not visible in the video though.
The video is packaged as a segment from a news broadcast but there is no visible logo. Captions in Hindi, offering transcription of the audio appear at the bottom of the video frame. There are several jump-cuts in the segment featuring the doctor.
Visuals of both Sharma and Chhajer are out-of-sync with their respective audio tracks. In the segments featuring the two men, it appears as if their chins extend beyond their natural contours when their lips move to talk. The doctor’s teeth seem to change shape and his cheeks appear unnaturally stiff.
The voice being attributed to Sharma sounds like his and so does the accent when compared with his recorded videos available online, however, the pauses characteristic of his delivery are missing.
The audio track with Chhajer’s visuals captured his voice’s raspiness and accent, but sounds scripted and monotonic. The track purportedly assigned to the police officer sounds very robotic with an unnaturally heavy and reverbed tone, it is hastened with no natural pauses.
Unlike Chhajer, there is no public statement from Sharma or the police officer with regard to this video.
Reverse image search using screenshots from the video led us to three different videos —first one, second one, and a third — featuring the three men seen in the video. This helped establish that the video we reviewed is composed of unrelated clips that have been stitched together.
The video featuring Sharma was published from the official YouTube channel of IndiaTV, a Hindi television channel, on Oct. 2, 2024. Chhajer’s video was published from the official YouTube channel of GetsetflySCIENCE (sic) by Gaurav Thakur on July 20, 2024. The police officer’s video was embedded with a tweet posted from the micro-blogging site X on Sep. 28, 2024.
The audio content in each of the videos is different from that in the doctored video. However, the clothing, backdrop, and body language of the subjects in these videos and their respective segments in the doctored video are identical. The logos and supers visible in the original videos are absent in the doctored video. The frames in the manipulated clips are more zoomed in.
To discern the extent of A.I.-manipulation in the video under review, we put it through A.I.-detection tools.
The voice tool of Hiya, a company that specialises in artificial intelligence solutions for voice safety returned results indicating that there is a 97 percent probability that an A.I.-generated audio track was used in the video.
Hive AI’s deepfake video detection tool indicated that the video was indeed manipulated using A.I. It pointed out many markers in the video, but only on Chhajer’s face. Their audio tool indicated that the majority of the audio track was manipulated with A.I., but for the last 20-second segment.
The deepfake detector of our partner TrueMedia suggested substantial evidence of manipulation in the video. The “A.I.-generated insights” offered by the tool provide additional contextual analysis by stating that the audio transcript reads like a promotional pitch, and uses tactics seen commonly in advertisements.
In a breakdown of the overall analysis, the tool gave an 81 percent confidence score to the “video facial analysis” subcategory that analyses video frames for unusual patterns and discrepancies in facial features. It also gave a 61 percent confidence score to the “face manipulation detector” subcategory that detects potential A.I. manipulation of faces, as in the case of face swaps and face reenactment.
The tool gave a 100 percent confidence score each to the subcategories of “AI-generated audio detector” and “voice anti-spoofing analysis”, which detect A.I.-generated audio. The tool also gave a 97 percent confidence score to the “audio authenticity detector” subcategory which analyses audio for evidence that it was created by an A.I. generator or cloning.
We reached out to ElevenLabs, a company specialising in voice A.I. research and deployment for an expert analysis on the audio. They told us that they were unable to confirm that the audio was A.I.-generated. The company has been actively identifying and blocking attempts to generate prohibited content, however, the exact generation of this audio from their platform has not been confirmed.
To get another expert to weigh in on the audio, we escalated it to our partner Validia, a San-Francisco based deepfake cybersecurity service. They used their proprietary software to check the authenticity of the voice associated with Sharma since that sounded very much like his voice.
They took two real voice samples of Sharma’s to generate an audio profile of his which served as a representation of his real voice. Then they isolated the voice from the video escalated by us, and removed the background audio to get a clean sample. Following which they used a heat-map analysis to compare the generated audio profile and the retrieved audio.
The image on the left displays the comparison between the two real voice samples of Sharma’s. The image on the right draws a comparison between a real voice sample and the retrieved audio from the doctored video.
The team at Validia stated that the deepfake audio is very similar to Sharma’s real voice. However, they added that there were slight differences which indicated that the audio was very likely generated as an A.I. clone of Sharma’s.
They also analysed the video track using their new proprietary video detection software, which identified very high levels of facial forgery in the video, and indicated the use of lip-sync techniques in generating the video.
To get another expert to weigh in on manipulations in the video, we reached out to our partners at RIT’s DeFake project. Kelly Wu from the project shared her observations on the various discrepancies evident in the video.
Referring to Sharma’s segment in the video, Ms. Wu noted that there is a gap in the audio track for a few seconds but despite that subtle movement can be observed in his mouth region. His lips appear to be moving even as there is no corresponding audio.
Wu added that in the police officer’s interview, despite his face being redacted his mouth movements are still somewhat discernible through the motions of the mouth region. Referring to a two-second clip in that particular segment, she noted that his mouth appears to be moving despite a gap in the audio.
She pointed to the presence of artefacts around Chhajer’s teeth. Suggesting that, for about three seconds his upper teeth seem to disappear intermittently while he appears to be talking. She added that at a certain point in the video his upper teeth appear as a distorted, squiggly white line.
On the basis of our findings and analyses from experts, we can conclude that the video of Sharma and Chhajer is fake. Synthetic audio seems to have been used over disparate clips of the subjects to spread misinformation about Chhajer’s death and peddle a suspicious cure.
(Written by Debraj Sarkar and Debopriya Bhattacharya, edited by Pamposh Raina.)
Kindly Note: The manipulated audio/video files that we receive on our tipline are not embedded in our assessment reports because we do not intend to contribute to their virality.
You can read below the fact-checks related to this piece published by our partners:
Viral Clip Of Rajat Sharma Reporting About Dr Bimal Chhajer's Death Made With AI Voice Clones
