16 August 2026

AI-manipulated videos of public figures announcing Wasim Akram’s death go viral

Account posting doctored videos about Wasim Akram’s death traced to TikTok

By Haseem uz Zaman

Claim: Former Pakistani cricketer Wasim Akram has died, according to video statements issued by multiple public figures.

Fact: Wasim Akram is not dead, he and his wife have confirmed. The viral clips of public figures are manipulated with artificial intelligence (AI) tools, according to analyses by different deepfake detectors and a sound engineer as well as the identification of the source videos.

In July 2026, videos featuring a number of public figures and media personalities announcing the death of cricket legend Wasim Akram went viral on social media platforms.

These include cricketers Shoaib Akhtar, Naseem Shah, Shaheen Shah Afridi, Shahid Afridi, Babar Azam, and Mohammad Rizwan, politician and journalist Iqrar ul Hassan, Pakistan Tehreek-e-Insaf (PTI) leader Shahbaz Gill, Islamic preacher Maulana Tariq Jameel, political analyst Irshad Bhatti, talk show host Mubasher Lucman, and Geo News anchors Faraz Ahmed and Raza Zaidi.

The announcements start with stating Akram’s name and expressing regret, before ending by saying that his death was caused by a heart attack, his funeral is scheduled for the same day, and asking for prayers for him.

The archived links of these viral videos can be found here, here, here, here, here, here, here, here, here, here, here, here, here, here, and here.

Fact or Fiction?

Soch Fact Check found that Wasim Akram (archive) and his wife, Shaniera (archive), have refuted the claim about the cricketer’s death.

“Thanks for the concern, I’m fine, thank you,” Akram said in a video statement posted on his Instagram account. He also questioned how people could fall for fake news.

In her Instagram post, Shaniera Akram wrote, “In response to the FAKE news and AI videos circulating , I just took this video for you and as you can see he is in the gym and seems he ok!”

Deepfake detection tools

Soch Fact Check traced most of the original videos and ran the viral clips through various deepfake and AI manipulation detectors, such as DeepFake-O-Meter, NVIDIA Synthetic Video Detector, Hive Detect, and Global Online Deepfake Detection System (GODDS).

We also sought a comment from Shaur Azher, a lecturer who teaches sound design and sound recording at the University of Karachi and the Shaheed Zulfikar Ali Bhutto Institute of Science and Technology (SZABIST). He works as an audio engineer at our sister organisation, Soch Videos, and specialises in mixing and mastering audio.

For the readers’ convenience, the analysis on each video is organised under a separate subheading, based on the name of the public figure appearing in the clip.

Shoaib Akhtar

Soch Fact Check traced the original video (archive) to Akhtar’s X (formerly Twitter) account, where it was posted on 9 March 2025.

According to DeepFake-O-Meter, the video is likely AI-generated, with “moderate” confidence. The detectors we used provided probabilities of 98.6%, 98.3%, 97.8%, 96.2%, 89.2%, 84.7%, 61.8%, and 46.2%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 64.2%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 39.1% likely to be AI-generated video, 1% likely to have AI-generated speech, 0% likely to contain AI-generated music, and 0% likely to be a deepfake. At the four-second mark, the clip’s likelihood of being AI-generated is 86.2%; specifically, it identified the chances of sora2 and seedance2 being the source of 49.4% and 24.1%, respectively.

GODDS — the tool developed by Northwestern University’s Security & AI Lab (NSAIL) that uses a combination of various models along with human analysis to provide a holistic summary of the results — used 22 deepfake detection algorithms for the visual content and 70 for the audio component, while two trained analysts also examined the clip.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 16 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the six other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 36 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 34 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

“The artefacts are largely obscured by poor video quality. Hence, it is difficult to discern possible artefact manipulation from the video’s heavy compression. These indicators can plausibly be explainable by other factors such as motion blur as well,” they said, adding that the sample “shares the same visual component as another video, which was posted in 2025”.

“We believe this media is likely manipulated” using AI, the analysts concluded.

Azher, the sound engineer, shared his findings, saying, “The data reveals distinct structural divergences between the tracks despite similar spectral averages. While the AI-processed track (Sample A) can closely mimic breath ratios and maintain a clean harmonics-to-noise ratio (HNR), the mathematical perfection in phase alignment (1.000), a highly monotone delivery (1.80 standard tone (ST)), and the unusually high amplitude instability (Shimmer at 82%) act as strong identifiers of synthetic origin or heavy digital processing compared to the mono baseline.”

Shahbaz Gill

Soch Fact Check traced the original video (archive) to Gill’s YouTube channel, where it was uploaded on 4 July 2026.

According to DeepFake-O-Meter, there were mixed signals about the video, with “low confidence”. The detectors we used provided probabilities of 99.9%, 99.7%, 97.3%, 90.7%, 74.8%, 37.4%, 22.6%, and 0%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 21.7%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 8.52-second mark, the score went up to 44%.

According to Hive Detect’s results, the sample is 11.9% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 1.3% Likely to be a Deepfake.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 10 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the 12 other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 55 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 15 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

They said, “The artefacts are largely obscured by poor video quality. Hence, it is difficult to discern possible artifact manipulation from the video’s heavy compression. These indicators can plausibly be explainable by other factors such as motion blur as well.”

“Given the subject’s political status, it is likely that there would be greater news coverage were this media authentic. The account that posted the video, 2jdjfjcsv61, is not an official account, and claims to not be affiliated with any person or organisation,” they added, concluding that the sample is “likely manipulated” using AI.

According to Azher, “The data presents compelling forensic evidence that Sample A is an AI-generated synthetic output, whereas Sample B serves as a natural, biological baseline.

“The most distinct indicator of synthesis in Sample A is the severe suppression of human respiratory dynamics; it yields a significantly lower percentage of breath frames and minimal biological breath events compared to the organic recording.

“Spectral analysis further supports this, with Sample A demonstrating a narrower MFCC [Mel-Frequency Cepstral Coefficients] standard deviation (10.70 vs 11.62), a hallmark of the algorithmic timbral smoothing typical in neural text-to-speech [TTS] models,” he concluded.

Naseem Shah’s first video

Soch Fact Check traced the original video (archive) to the YouTube channel The SB Cricket, which posted it on 23 December 2019.

According to DeepFake-O-Meter, there were mixed signals about the video, with low confidence. The detectors we used provided probabilities of 100%, 100%, 99.8%, 99.4%, 69.9%, 54.1%, 12.8%, and 0%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 29.7%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 6.05-second mark, the score went up to 64.3%

According to Hive Detect’s results, the sample is 12.2% Likely to be AI-Generated Video, 37.5% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0% Likely to be a Deepfake. It identified a 39% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to five of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the 17 other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 44 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 26 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

They said, “The artifacts are largely obscured by poor video quality. Hence, it is difficult to discern possible artifact manipulation from the video’s heavy compression. These indicators can plausibly be explainable by other factors such as motion blur as well.”

The clip “shares the same visual component as another video, which was posted in 2019”, the analysts added, saying it suggested “the possibility of audio manipulation based on the original one”. They “believe this media is likely manipulated” using AI.

In this regard, Azher said, “The data reveals quantifiable structural divergences between Samples A and B.

“Sample A exhibits notably higher jitter (16.597% vs 10.979%) and shimmer (59.163% vs 52.962%) compared to Sample B. In acoustic forensics, this elevated frequency and amplitude instability is a primary indicator of increased vocal instability or potential synthetic generation artifacts in NS01.”

Naseem Shah’s second video

Soch Fact Check traced the original video (archive) to the Pakistan Cricket Board’s (PCB) YouTube channel, where it was posted on 3 November 2025.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 100%, 97%, 83.8%, 37.5%, 25.4%, 10.4%, and 0.1%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 32.5%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 50.7% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0.6% Likely to be a Deepfake. It identified a 78.2% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 12 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the 10 other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 31 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 39 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The sample “shares the same visual component as another video, which was posted in 2025”, the analysts said, noting that it suggested “the possibility of audio manipulation based on the original video”. They concluded that the clip “is likely manipulated” using AI.

Azher explained that the pattern is “consistent with real, unprocessed human speech, and notably consistent with genuine emotional strain in the voice”.

“Paired with the independently confirmed news reporting that the Wasim Akram death story was an AI-generated deepfake using fabricated cricketer voices.”

Iqrar ul Hassan

Soch Fact Check traced the original video (archive) to Hassan’s YouTube channel, where it was uploaded on 6 June 2026.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 100%, 89.3%, 60.8%, 41.6%, 28.5%, 8.6%, and 2.2%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 36.3%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 0.1% Likely to be AI-Generated Video, 3.7% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0.1% Likely to be a Deepfake. It identified a 28% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 17 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the five other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 48 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 22 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

They concluded that the video is “likely manipulated” using AI.

Azher explained his findings, saying, “The data indicates measurable structural divergences between Sample A and Sample B. Most notably, Sample A exhibits mathematical perfection in phase alignment (1.000) — a metric frequently associated with neural vocoder synthesis — despite its lower overall mean coherence (0.828).

“In contrast, Sample A displays higher mean coherence (0.956) but imperfect phase alignment (0.893), suggesting a more organic or standard digital capture process. The jitter and shimmer metrics are relatively close, indicating that Sample A’s primary indicator of artificiality is the phase alignment score, whereas Sample B presents a more balanced, albeit distinct, acoustic profile.”

Shaheen Shah Afridi

Soch Fact Check traced the original video (archive) to New Zealand Cricket’s (NZC) official YouTube channel, where it was posted on 17 January 2024.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 99.1%, 56.8%, 46.1%, 24.5%, 19.3%, 1.3%, and 0.3%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 19.2%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 12.66-second mark, the score went up to 59.5%.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0% Likely to be a Deepfake. It identified a 52.1% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 17 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the five other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 23 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 47 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The sample “shares the same visual component as another video, which was posted in January of 2024”, the analysts noted, saying it suggested “the possibility of audio manipulation based on the original video”. They concluded that the clip is “likely manipulated” using AI.

Azher stated, “The comparative analysis of Samples A and B reveals measurable technical divergences in their acoustic signatures. While both files show consistent, high quality stereo phase alignment, the variations in MFCC distribution, breath heuristics, and micro variations (jitter and shimmer) suggest distinct differences in recording environments or post production processing paths.

“Notably, Sample B demonstrates a broader timbral variance and higher breath frequency, whereas Sample A shows slightly elevated jitter and shimmer metrics, which are often indicative of localized artifacting or synthetic generation patterns when compared against a more natural baseline.”

Irshad Bhatti

Soch Fact Check traced the original video (archive) to Geo News’ YouTube channel, where it was posted on 27 June 2026.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 99.4%, 92.7%, 74.6%, 61%, 36.2%, 20.3%, and 0.3%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 34.7%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0.1% Likely to be a Deepfake. It identified a 1% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 18 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the four other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 35 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 35 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The analysts explained, “The artifacts are largely obscured by poor video quality. Hence, it is difficult to discern possible artifact manipulation from the video’s heavy compression. These indicators can plausibly be explainable by other factors such as motion blur as well.”

The video “appears to be taken from a Geo News interview, due to the logo in the video”, they added, concluding that it is “likely manipulated” using AI.

In this regard, Azher said, “The comparative analysis shows measurable differences in acoustic stability between Samples A and B. The former exhibits lower frequency instability (7.876% vs 11.020% jitter) and lower amplitude instability (24.668% vs 35.646% shimmer).

“Notably, Sample A maintains a cleaner voice quality profile, while Sample B has a significantly higher total cycle count. The absence of detected breath events in some segments across the extended duration (69.6 seconds), compared to the higher event in the shorter Sample A, suggests distinct recording or normalization variances.

“Overall, the lower micro-variation metrics (jitter and shimmer) in IB01 suggest a more stable, higher-fidelity acoustic profile compared to Sample B which proves Sample A is AI-generated.”

Shahid Afridi

Soch Fact Check traced the original video (archive) to the Pakistani marketing and communications services company The Nutshell Group’s YouTube channel, where it was posted on 22 September 2017.

According to DeepFake-O-Meter, there were mixed signals about the video, with low confidence. The detectors we used provided probabilities of 100%, 99.9%, 99.6%, 98.3%, 97%, 48%, 46.1%, and 13%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 37.4%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 1% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0.3% Likely to be a Deepfake. It identified a 1% chance of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 17 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the five other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 43 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 27 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated using AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The sample “shares the same visual component as another video, which was posted in 2017”, the analysts explained. “This suggests the possibility of audio manipulation based on the original video,” they said, concluding that the clip is “likely manipulated” using AI.

Azher shared his observations, stating, “The data reveals structural divergences between the two audio sources. While both samples share suspiciously perfect phase alignment scores and high amplitude instability (Shimmer > 55%), Sample B exhibited MFCC standard deviation variance and a higher prevalence of biological breath markers.

“Taken together, localized indicators (fewer breath events, narrower timbral distribution, and lower jitter) point toward heightened audio processing patterns or synthetic generation in Sample A.”

Babar Azam’s first video

Soch Fact Check traced the original video (archive) to the PCB’s Facebook page, where it was posted on 4 November 2023.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 99.9%, 99.7%, 97.8%, 77.1%, 59.5%, 34%, 14.4%, and 10.4%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 12.5%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 0.03-second mark, the score went up to 54.6%.

According to Hive Detect’s results, 0% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 2.7% Likely to be a Deepfake. At the eight-second mark, the likelihood of the clip being a Deepfake is 73.6%.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 14 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the eight other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 47 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 23 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

“The original content uploader of this media, ‘Fazal Nabi’, has posted numerous videos that appear to be AI-generated without independent corroboration or cross-references. This pattern raises concern about the account’s reliability,” they explained, concluding that the sample is “likely manipulated” using AI.

According to Azher, “Sample A is conclusively classified as AI-generated speech, while Sample B represents authentic human speech. This distinction is most strongly evidenced by Sample A’s mathematically artificial 1.000 stereo phase alignment, which exposes synthetic mono-to-stereo duplication, lacking real acoustic space alongside its higher average MFCC standard deviation (9.52) and dramatically elevated shimmer (58.46%), both of which reveal the spectral envelope variance and cycle-to-cycle amplitude instability typical of neural vocoders.

“Conversely, Sample B exhibits the biological and physical realities of a genuine microphone recording, maintaining lower amplitude instability (49.95% shimmer), natural inter-channel phase variations (0.995 alignment), and a higher proportion of respiratory breath frames (18.83%), which generative models consistently suppress or fail to organically replicate.”

Babar Azam’s second video

Soch Fact Check traced the original video (A) to the PCB’s YouTube channel, where it was posted on 6 July 2026.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 92.1%, 79.1%, 37.6%, 16%, 12.9%, 0.3%, and 0.1%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 25.2%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 6.19-second mark, the score went up to 70.6%.

According to Hive Detect’s results, the sample is 0.8% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0% Likely to be a Deepfake.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 17 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the five other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 37 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 33 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

“This artifact shares the same visual component as another video, which was posted on July 6th,” the analysts noted, saying it suggested “the possibility of audio manipulation based on the original video”. Therefore, the clip is “likely manipulated” using AI, they concluded.

Azher explained his analysis, noting that while “AI can closely mimic frequency instability (Jitter), the mathematical perfection in phase alignment and the unusually high amplitude instability (Shimmer) act as strong identifiers of synthetic origin in sample Sample A”.

“Conversely, the human sample is characterized by natural phase variance and a higher prevalence of biological breath markers in Sample B.”

Mohammad Rizwan

Soch Fact Check traced the original video (archive) to the PCB’s YouTube channel, where it was posted on 7 March 2026.

According to DeepFake-O-Meter, there were mixed signals, with low confidence. The detectors we used provided probabilities of 99.9%, 92.8%, 42.7%, 10.9%, 9.4%, 8.8%, 8.4%, and 1.1%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 17.4%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 4.85-second mark, the score went up to 56.6%.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 25.3% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0% Likely to be a Deepfake.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 15 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the seven other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 49 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 21 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

“The user who posted the video, Khalid Baloch, is not an official account and claims to not be affiliated with any person or organization,” they said. They also observed that the sample “shares the same visual component as another video, which was posted in March of 2026”, suggesting the “possibility of audio manipulation based on the original video”.

The clip is “likely manipulated” using AI, they concluded.

Azher, the sound engineer, said, “The data indicates measurable structural divergences between Samples A and B. Notably, Sample A displays an imperfect phase alignment (0.970) compared to MR02’s near perfect alignment (0.999). While this is an inversion of the mathematically perfect phase pattern often observed in synthetic audio, it nonetheless indicates a significant divergence in the mastering or generation environment. Sample A exhibits lower frequency instability (jitter: 5.523% vs 9.134%) but maintains comparable amplitude instability (shimmer) levels to Sample B.

“Combined with the narrower MFCC variance, these indicators suggest that Sample A may have undergone distinct post-processing or synthesis techniques that differentiate it from Sample B’s baseline.”

Maulana Tariq Jameel

Soch Fact Check traced the original video (archive) to the YouTube channel Hanafi Scholars, where it was posted on 6 February 2021.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 98.7%, 84.4%, 82.4%, 59.6%, 36.9%, 25.2%, and 6.7%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 26.2%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 8.39-second mark, the score went up to 47.2%.

According to Hive Detect’s results, the sample is 0.8% Likely to be AI-Generated Video, 0% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 5.9% Likely to be a Deepfake.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 12 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the 10 other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 54 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 16 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated using AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The clip is “likely manipulated” using AI, they concluded.

Azher said the two samples “show the largest MFCC Euclidean distance (108.220) in the current batch”. The data, he added, “presents an unusual phase pattern: Sample A maintains extremely high coherence (0.999) but imperfect alignment (0.991), whereas Sample B exhibits perfect alignment (1.000) but significantly lower coherence (0.907).

“Furthermore, Sample B shows notably higher jitter and shimmer metrics compared to Sample A. While higher jitter/shimmer typically suggests natural instability, the mathematical perfection in phase alignment often points toward synthetic processing. The conflicting technical indicators suggest these signals do not align cleanly in one direction, though the evidence leans toward Sample A containing more heavily processed or synthetic characteristics.”

Mubasher Lucman

While Soch Fact Check could not trace the original video, we observed that Lucman’s shirt features the word “Baaghi,” which likely refers to Baaghi TV, where he serves as the chief executive officer (CEO) and an anchor. For comparison purposes, we used a video from the outlet’s YouTube channel that was uploaded on 29 April 2023 and features the same setting, such as the bookshelf, sofa, and the door in the background.

According to DeepFake-O-Meter, there were mixed signals, with low confidence. The detectors we used provided probabilities of 100%, 99.9%, 99.8%, 99.5%, 96.9%, 50.9%, 44.3%, and 37.7%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 6.1%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 5.58-second mark, the score went up to 36.9%.

According to Hive Detect’s results, 0% Likely to be AI-Generated Video, 92.1% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0% Likely to be a Deepfake. It identified a 92.1% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 16 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the six other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 42 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 28 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The clip is “likely manipulated” using AI, they concluded.

In this regard, Azher observed, “Both samples demonstrate perfect (1.000) phase alignment. However, Sample A exhibits notably higher frequency instability (jitter: 13.942% vs 9.617%) and amplitude instability (shimmer: 64.077% vs 56.144%). Furthermore, the breath heuristic flags a significantly higher percentage of frames in Sample A (15.33%) compared to Sample B (1.57%).

“Combined with Sample A’s more monotone delivery and elevated shimmer levels, these indicators lean toward it exhibiting more heavily processed or synthetic audio characteristics compared to Sample B.”

Faraz Ahmed

While Soch Fact Check could not trace the original video, we observed that Ahmed has worn the same tie multiple times — on 7 July 2026, 15 May 2026, 21 February 2026, 20 February 2026, 19 February 2026, 8 December 2025, and 26 June 2025 — and used one for comparison purposes.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 99.6%, 99.6%, 98.7%, 88.2%, 51.6%, 49.8%, 13.9%, and 6.6%.

NVIDIA Synthetic Video Detector said the sample is “synthetic”, with a score of 37.5%, which is “greater than threshold (30%), meaning it’s likely synthetic or AI-generated”.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 5.5% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 1.1% Likely to be a Deepfake. At the 12-second mark, the likelihood of the clip being a deepfake is 56.5%.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 14 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the eight other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 49 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 21 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The video is “likely manipulated” using AI, they concluded.

Azher said, “Sample A maintains a more monotone, cleaner vocal profile with higher HNR and lower perceived shimmer, suggesting a more stable or processed delivery. Conversely, Sample B, despite maintaining a slightly higher mean coherence in stereo phase, exhibits higher relative shimmer in the micro-variation analysis and a lower HNR, which often correlates with a more organic or breathy acoustic capture, indicating that Sample A is cloned or generated using AI.”

Raza Zaidi

Soch Fact Check traced the original video on the Internet Archive’s Wayback Machine, where it was uploaded on 4 July 2026.

According to DeepFake-O-Meter, the results were inconclusive, with very low confidence. The detectors we used provided probabilities of 100%, 100%, 79.9%, 49.8%, 36.2%, 35.4%, 4.2%, and 0.2%.

NVIDIA Synthetic Video Detector said the sample is “real”, with a score of 26.7%, which is “lower than threshold (30%), meaning it’s likely a real video”. At the 2.31-second mark, the score went up to 51.8%.

According to Hive Detect’s results, the sample is 0% Likely to be AI-Generated Video, 0.4% Likely to have AI-Generated Speech, 0% Likely to contain AI-Generated Music, and 0.1% Likely to be a Deepfake. It identified a 12% likelihood of AI-Generated Speech.

All predictive models for the visual and audio content said the video “is likely to be fake”:

The video is likely to be fake with a probability above 0.5, according to 16 of the 22 predictive models; it is likely to be fake with a probability below 0.5, according to the six other predictive models.
The audio is likely to be fake with a probability above 0.5, according to 48 of the 70 predictive models; it is likely to be fake with a probability below 0.5, according to the 22 remaining predictive models.

According to GODDS’ human analysts, the video contains “several indicators” that show it may be artificially manipulated via AI. For example:

The voice does not match the natural cadence and breathing patterns of genuine speech.
The lip movements do not align with the sounds.

The clip is “likely manipulated” using AI, they concluded.

Azher explained his findings, saying, “Sample A exhibits two major red flags commonly associated with AI generation: mathematical perfection in phase alignment (1.000) and extreme amplitude instability. The shimmer value in Sample A reaches 86.089%, heavily exceeding the >55% threshold typical of neural vocoders struggling to perfectly reconstruct amplitude stability across analyzed cycles.”

Therefore, we conclude that the viral clips are AI-generated and that Akram is alive.

Virality

Soch Fact Check found the videos posted multiple times on different social media platforms, especially TikTok, where they gained hundreds of thousands of views.

Interestingly, we noted that the main TikTok account — which appears to be the source of the fake news — has changed its handle from @2jdjfjcsv6s to @syatt480.

Conclusion: Wasim Akram is not dead, he and his wife have confirmed. The viral clips of public figures announcing his death are manipulated using AI tools, according to analyses by different deepfake detectors and a sound engineer, as well as the identification of the source videos.

Cover photo: Soch Fact Check

To appeal against our fact-check, please send an email to appeals@sochfactcheck.com