Deepfakes drastically improved in 2025. They're about to get even harder to detect

"Over the course of 2025, deepfakes improved dramatically. AI-generated faces, voices, and full-body performances that mimic real people increased in quality far beyond what even many experts expected would be the case just a few years ago. They were also increasingly used to deceive people. For many everyday scenarios-especially low-resolution video calls and media shared on social media platforms-their realism is now high enough to reliably fool nonexpert viewers."

"And this surge is not limited to quality. The volume of deepfakes has grown explosively: Cybersecurity firm DeepStrike estimates an increase from roughly 500,000 online deepfakes in 2023 to about 8 million in 2025, with annual growth nearing 900%. I'm a computer scientist who researches deepfakes and other synthetic media. From my vantage point, I see that the situation is likely to get worse in 2026 as deepfakes become synthetic performers capable of reacting to people in real time."

"First, video realism made a significant leap, thanks to video generation models designed specifically to maintain temporal consistency. These models produce videos that have coherent motion, consistent identities of the people portrayed, and content that makes sense from one frame to the next. The models disentangle the information related to representing a person's identity from the information about motion so that the same motion can be mapped to different identities, or the same identity can have multiple types of motions."

During 2025, deepfakes improved substantially in visual and audio fidelity and were increasingly used to deceive people. Realism became sufficient to reliably fool nonexpert viewers in many everyday scenarios such as low-resolution video calls and social media media. Synthetic media became indistinguishable from authentic recordings for ordinary people and sometimes institutions. The volume of online deepfakes surged massively, rising from roughly 500,000 in 2023 to about 8 million in 2025. Technical advances produced temporally consistent video generation models that disentangle identity from motion, producing stable, coherent faces without the former artifacts that aided forensic detection.

Read at Fast Company