Between 25 November and 2 December, several Russian-language Facebook accounts published a video (1; 2; 3) showing an interview with a woman speaking Ukrainian. In the video, she says she opposes a peace agreement and argues that the war should continue until Ukraine’s territory is fully liberated. When asked about her own family’s participation in the war, the woman replies that her son is in Warsaw and her daughter is in Germany. She emphasizes that both of her children spiritually support the soldiers fighting in Ukraine. The text displayed in the video reads, “We are Ukrainian citizens living in Europe, and we demand the restoration of the 1991 borders. No peace negotiations with Russia. We demand the continuation of the banquet.”
The circulated video in which a Ukrainian woman calls for the continuation of the war is false. Several visual inconsistencies, together with AI-detection tools, confirm that both the video and the audio of the interview were generated using artificial intelligence.
We identified several inconsistencies in one of the frames. Specifically, the letters printed on the lower part of the microphone change several times as the microphone moves throughout the video, while the text on the red section is unreadable – its letters appear blurry and indistinct. The lower part of the microphone, which changes multiple times, bears the word “НОВИНИ,” which means “news” in Ukrainian. We attempted to find information about any media outlet that uses a microphone with such a design, but we were unable to locate one in open sources.
The design on the woman’s shirt also appears unnatural and, upon closer observation, seems to move. Additionally, the video and audio are not synchronized – the woman’s voice can be heard before any mouth movement is visible. This mismatch between voice and mouth movement is one of the characteristic signs of AI-generated footage.
We further analyzed the video using verification tools. Hiya, an InVID tool designed to detect AI-generated voice cloning, confirms with 99% accuracy that the voice is AI-generated. According to the analysis of another AI-detection tool, Hive Moderation, there is a 99.8% probability that the video contains AI-generated content. The tool also detected the use of artificial intelligence in the audio segment with a high degree of accuracy.
The article has been written in the framework of Facebook’s fact-checking program. You can read more about the restrictions that Facebook may impose based on this article via this link. You can find information about appealing or editing our assessment via this link
Read detailed instructions for editing the article.
Read detailed appeal instructions.
