Video of Pakistan Army soldiers praising their commander was generated by AI

Claim: A video shows soldiers praising their commanders during recent attacks across Balochistan.

fact: Viral clips may have been generated by AI.

Following the recent wave of deadly attacks across Balochistan; clip pakistani army soldier Words of praise for his command were shared online. The video, which shows a soldier acknowledging that the “mission” was accomplished, was shared with the caption:

“ کیلئے دعا کریں

اللہ تعالیٰ آپ سب کو اپنی حفظ امان میں رکھے آمین ثم آمین یا رب العالمین

پاک افواج زندہ باد ❤️❤️❤️

Oh, thank you.”

[Translation: Commander Sahib was ahead of now. He has been shot. Pray for his health and life.

May Allah Almighty keep you all in His protection. Amen and Amen, O Lord of the Worlds.

Long live Pak Army ❤️❤️❤️

Islam Pakistan]

Attack in Balochistan

On January 31, 2026, several police stations in Quetta, Balochistan targeted According to reports, the attack, which began around 3 a.m. local time, was carried out by an armed group. Al Jazeera. “Grenade and attack with gun Pakistan’s Deputy Interior Minister Talal Chaudhry said hospitals, schools, banks and markets were targeted in the provincial capital Quetta and other cities and towns across the province. BBC Reported. “In both cases, the attackers came disguised as civilians and indiscriminately targeted civilians working in shops,” he said, adding that the attackers used civilians as human shields. report Added.

” killed in attack The military said at least 15 security personnel and 18 civilians were participating. Officials said 92 militants were also killed, including “three suicide bombers.” france 24. of report It added: “The dead in Balochistan’s Gwadar city included 18 civilians, including three women and three children, police official Ibad Khan said.” said Police quickly responded to the attack and killed all of the assailants. ”

of banned separatist groups Balochistan Liberation Army (BLA) claims responsibility Some banks were also robbed during the attack. Pakistan’s Balochistan Chief Minister Sarfaraz Bugti said: security forces killed A 40-hour battle between 145 insurgents. 17 Law He added that as of February 1, 31 enforcement personnel and civilians had been killed in militant attacks.

Fact or fiction?

Soch fact check I have debunked a similar hoax before. video of soldier. This viral clip contains multiple visual inconsistencies, suggesting it was likely generated using an AI text-to-video model.

Significant signs are static background. While the soldier is talking to the camera, the soldier sitting behind him is completely motionless. Doesn’t movethere are two other soldiers on the left side of the screen. In contrast, two other soldiers can be seen walking in the background. This natural movement mismatch is a common indicator in AI-generated videos where background elements often cannot be animated realistically.

A closer look at the soldiers’ uniforms revealed other signs of AI. The patch on his left arm contains no readable name, logo, or designation. Instead, the text appears garbled. AI-generated content occurs frequently conflict It renders text correctly, especially on clothing, badges, and signs.

Visual contradictions in viral videos

Additionally, photos of official uniforms available online from news sites and social media show that soldiers’ combat vests typically do not carry nametags or Pakistani flags. In fact, the uniforms of Pakistani soldiers usually display “Pak Army” along with the wearer’s name, and a flag patch is usually seen on the arm.

pakistan army uniform

Importantly, the reel included a logo that read “Frontline Vilogs,” indicating it may have been manipulated or edited. we, Frontline video blog Found on Facebook page multiple resemble video This account also appears to have been generated by AI.

Soch fact check I then tested the video Deepfake-O-meteranalyzed using multiple AI-based detection models. The video results are below.

DeepFake-O-Meter results

We first used the AVSRDD (2025) model. This is an AVSR-based audio and visual deepfake detection method that leverages audio correlation. This model uses dual branch encoders for audio and video to support independent detection of each modality. The video was rated 100% likely to be fake.

Next, we used the DSP-FWA (2019) model. We focus on detecting deepfakes by identifying facial distortion artifacts introduced during the deepfake generation process. The probability that the video is fake was estimated at 97.7%. This model uses digital signal processing (DSP) techniques to identify inconsistencies that occur when a synthesized face is resized and blended into the original image or video.

This video was analyzed using LSDA (2024), a deepfake detection model designed to assess whether a video or image is fully or partially synthesized using AI. The model evaluates visual and temporal cues such as facial movement, lip synchronization, and texture consistency to estimate the likelihood of a synthesis operation. The probability that this clip is fake was rated at 54.5%.

Finally, we used the WAV2LIP-STA (2022) detection model. It’s designed to detect lip-sync-based deepfakes, where mouth movements don’t exactly match the audio. This clip has a rating of 85.4%, indicating a high probability of tampering.