The Deepfakes Analysis Unit (DAU) analysed a video that apparently features Anant Ambani, non-executive director at Reliance Industries and their philanthropic arm, purportedly promoting an income-generating program. After putting the video through A.I. detection tools and getting our expert partners to weigh in, we were able to conclude that synthetic audio was used to fabricate the video.
A link to the Facebook post with the 42-second video embedded was sent to the DAU tipline for assessment. The video is in English and has garnered more than 220,000 views since it was uploaded on Dec. 19, 2024. As per the profile details of the account from which the video was posted, it belongs to a “gaming video creator”.
The video has been packaged as a short interview of Mr. Ambani filmed in a tight close-up, making only a part of the backdrop visible. A logo resembling that of the digital vertical of the Republic Media Network, an Indian news network, is visible in the top right of the video frame. The words “change your life today” can be seen in bold across the bottom of the video frame; the colour scheme for the text graphics emulates the red and white colour combination of the logo.
The tone of the video is bombastic. A male voice recorded over the video track makes a case for earning “passive income” through “immediate connect”, a supposed income generation program. It claims that the purported opportunity requires an initial investment of “21,200 rupees” and has already benefitted more than “60,000 users”. The video ends with the same voice urging people to register for the program and suggesting that it can improve India’s economic future.
Ambani’s lip movements seem mostly synchronised with the audio. His lips appear slightly blurred; and whenever he opens his mouth to speak it seems as if he has an extra set of teeth. The voice sounds somewhat like his and even attempts to mimic the natural pauses in human speech.
A distinct sound resembling that of laboured human breathing can be heard whenever there is a pause in the audio. The accent in the audio track sounds foreign when compared with Ambani’s recorded speeches and interviews available online. The overall delivery is hastened and scripted without much change in tone or pitch.
We undertook a reverse image search using screenshots from the video and found this video published on Feb. 26, 2024 from the official YouTube channel of ANI News, an Indian news agency. The clothing and head movements of Ambani in this video and the one we received on the tipline are identical, however, the framing of the two videos is different; making the backdrop look similar but not identical.
A zoomed in clip from the video we traced has been used to create the manipulated version. The ANI logo and the microphone visible in the original video have been cropped out. The logo associated with the Republic network and the text graphics are not part of the original video.
In August the DAU debunked a video peddling a fake gaming application, which featured Ambani along with other public figures. In that manipulated video too, clips from the same original video were used but the audio track was different just like the case in the video being addressed through this report. There is no mention of any income generating program in the original video, the language used is Hindi not English.
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, indicated that there is a 99 percent probability that the audio track in the video was generated or modified using A.I.
Hive AI’s deepfake video detection tool did not indicate A.I. manipulation in the video track. Their audio detection tool, however, traced A.I. manipulation in the entire audio track in the video.
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 video transcript reads like a promotional pitch for a financial program making bold claims about potential earnings. The insights also highlight that the mention of specific numbers and the use of the analogy of planting seeds to harvest crops is typical of marketing material rather than a genuine, unscripted conversation.
In a breakdown of the overall analysis, the tool gave a 63 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 subcategory of “video facial analysis” that checks video frames for unusual patterns and discrepancies in facial features, received a confidence score of 49 percent.
The tool gave a 100 percent confidence score to both “voice anti-spoofing analysis” and the “A.I.-generated audio detector” subcategories that help establish whether the audio is A.I.-generated. It gave a 99 percent confidence score to the “audio authenticity detector” subcategory, that analyses if the audio was generated by an A.I. audio-generator or cloning.
For a further analysis on the audio track we also put it through the A.I. speech classifier of ElevenLabs, a company specialising in voice A.I. research and deployment. The tool returned results indicating that it was very likely that the audio track used in the video was generated using their platform.
For expert analysis, we shared the video with our detection partner ConTrailsAI, a Bangalore-based startup with its own A.I. tools for detection of audio and video spoofs.
The team ran the video through audio as well as video detection models, the results that returned indicated high confidence for manipulation in the audio track, but not in the video track.
In their report they mentioned that the voice in the video track sounds monotonous and does not sound like Ambani’s. They noted that while their video model has predicted the video to be real overall, certain parts of the video have been identified as manipulated. They added that the results are ambiguous because Ambani’s face has been zoomed in on in the video.
To get another expert to weigh in on the video, we escalated it to the Global Deepfake Detection System (GODDS), a detection system set up by Northwestern University’s Security & AI Lab (NSAIL). They used a combination of 22 deepfake detection algorithms and analyses from two human analysts trained to detect deepfakes to review the video.
Of the 22 predictive models used to analyse the video, eight gave a higher probability of the video being fake and the remaining 14 indicated a lower probability of the video being fake.
The team observed that Ambani’s mouth and jaw seem to move in a more exaggerated manner than the rest of the face, contributing to an unnatural appearance. They added that the mouth appeared blurrier than the rest of the facial features.
They also pointed to specific frames where Ambani’s beard seems to overlap or intersect with the microphone, resulting in an unnatural appearance. Only a tiny portion of the microphone from the original video has been retained in the manipulated video. Their observations about Ambani’s voice corroborated our own. The team concluded that the video is likely fake.
On the basis of our findings and expert analysis, we can conclude that the video featuring Ambani was manipulated using A.I. voice. It is yet another instance of promoting a scam by linking it to a prominent business person.
(Written by Debraj Sarkar and Debopriya Bhattacharya, edited by Pamposh Raina.)
Kindly Note: The manipulated video/audio files that we receive on our tipline are not embedded in our assessment reports because we do not intend to contribute to their virality.
