How to identify that the “dark-skinned baby” suspicion video is AI-generated
A video has been posted from the official Facebook page of a Bangladeshi news outlet claiming that a husband is suspecting his wife because their child was born dark-skinned. Similar videos have also been posted by various Facebook users. A fact-check by Dismislab shows that the video posted by the Bangladeshi news outlet was created with the help of artificial intelligence (AI).
Earlier, multiple fact-checking organizations in Bangladesh and other countries published reports stating that similar videos were also produced using artificial intelligence. Despite verification by several fact-checking institutions, news outlets and social-media users continue to be misled. Experts say that although AI-generated videos are becoming more convincing with technological advancement, there are still ways to identify them. Even with new methods emerging, we must repeatedly return to the traditional questions of verification: Who? Where? Why?
A video (archive) bearing the caption “Because the baby is dark-skinned, the husband suspects his wife” was posted from the official Facebook page of the Bangladeshi news outlet Dainik Jaijaidin. The video’s hashtags read “Emotional Story” and “Family Drama.” In a room resembling a hospital cabin, a woman is lying down, holding two newborn babies in her arms and crying. A man is standing beside her. Shouting, he says, “What is happening right now? Are they my children? I don’t know. These are not my babies.”
At the 16-second mark, another scene appears in which the setting is the same, but the people have changed. Here too, a woman is seen holding two infants and crying, while a man beside her shouts, “What is happening right now? Are they my children? Oh my God! Oh my God! They have black hair, they have black skin. What is going on? They are not my babies.”
When attempting to verify the source of the videos, Dismislab identified a TikTok account. The account’s name is Braided. A video uploaded from Braided on October 31 matches exactly with the video circulating in Bangladesh. The caption of the TikTok video reads: “When the Babies Don’t Match the DNA 😱😂 Pure chaos in the delivery room! A young couple welcomes twins… but they look completely different from what anyone expected. The dad can’t believe it, the mom’s in tears, and the doctor’s just trying to keep it together. Real reactions, real confusion — straight out of a comedy movie!”
TikTok’s policy on “Creator labeled as AI-generated” states: “This label appears on content that a creator indicates was completely AI-generated or significantly edited with AI.” In other words, the video has been identified as AI-generated by its own creator.
When verifying the second video, Dismislab found a video report by the Indian fact-checking organization Factly. The video examined in that report matches exactly with the second segment of the video posted by Dainik Jaijaidin. According to Factly’s report, this video had circulated on social media as if it were real footage.
Some people even added new storylines, claiming that the woman had traveled to India before giving birth. The report states that the video had been posted from a TikTok account named Braided. However, during Dismislab’s verification, the video could no longer be found on that account. But the video could be found on a website called Urlebird. Urlebird is a third-party site that allows people to view TikTok videos, profiles, hashtags, and trends without logging into TikTok. In other words, it can be confirmed that this video too was posted from Braided.
Factly’s report states that after running the video through an AI-detection tool, it became even more certain that the footage had been created with the help of artificial intelligence.
An analysis of the Braided page shows multiple videos with nearly identical scripts and settings (1, 2, 3). The account also features an album of 27 videos titled “Delivery Room.”
When the same video spread earlier on social media, the fact-checking organization Boom Bangladesh investigated it. In its report, Boom Bangladesh also identified these videos as AI-generated. Despite the fact-check report being published on November 7, Dainik Jaijaidin posted the video on its Facebook page without any warning. They later removed it.
How to identify AI videos?
In a report published by Thomas Germain, a senior technology journalist at the BBC, several methods for identifying AI-generated videos are outlined. He writes, “AI videos are not more likely to look bad. The best AI tools can deliver beautiful, polished clips. And low-quality clips aren’t necessarily made by AI, either.”
In the same report, quoting Hany Farid, a computer science professor at the University of California, Berkeley, it is stated: “The leading text-to-video generators like [Google’s] Veo and OpenAI’s Sora still produce small inconsistencies. But it’s not six fingers or garbled text. It’s more subtle than that.”
The report states that three elements must be considered carefully: resolution (clarity), quality, and length. In most cases, AI-generated videos are very short—shorter than the short videos ((typically 30 to 60 seconds) we usually see on social media. The main reason is that generating video through AI is expensive. Moreover, the longer the video, the greater the likelihood that inconsistencies will become visible. As a result, there is a noticeable tendency to stitch together multiple short clips.
Regarding the other two components, the report explains that resolution and quality are related but not the same. Resolution refers to the number or size of pixels in an image, whereas compression reduces file size by discarding detailed information. This often produces block-like patterns and blurry edges. The purpose of this is to obscure fine details so viewers cannot clearly identify irregularities caused by the reduced video quality, making the footage appear more believable. This is a common tactic.
AI is improving continuously, and the possibility that subtle inconsistencies will no longer be visible to the naked eye is also increasing. Any new advice may become ineffective almost immediately due to the rapid introduction of new technologies. Thomas Germain says, “You’d never look at a piece of text and assume it’s true just because someone wrote it down. If there’s any question, you go investigate the source of the information. Videos and images used to be different, because they were harder to fake and manipulate. That’s over now. The only thing that matters now is where a piece of content came from, who posted it, what the context is and whether it’s been verified by a trustworthy source.”
