Expired50% confidencePredictionExact time
As the next generation of video generation models evolves, the window of opportunity for exposing fake faces through occlusion may not last long.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/18/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/16/2026(expired)
Sources
Related Claims
UnverifiedWith the maturation of open-source projects such as FaceSwap and DeepFaceLab, and the emergence of face-swapping tools based on Stable Diffusion, some tools can generate highly credible dynamic face-swapping videos with just a single photo.71% similarUnverified直接用场景加人物加提示词硬生成视频常遇到质量低下、画面穿帮、镜头跳脱、前后不接续等问题，在长镜头或多镜头叙事中尤为突出71% similarUnverified仅凭一张照片和一段录音即可生成高度逼真说话视频，降低了深度伪造的制作门槛67% similarUnverifiedThe video clips generated by Seedance are already very close to the standard of real human performance, with natural character movements and delicate visual quality.67% similarUnverifiedAround 2022, Meta's Make-A-Video and Google's Imagen Video first demonstrated the possibility of generating short videos directly from text, though the image quality was crude and the duration extremely short.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/552371API
curl https://kongchang.com/api/v1/knowledge/claims/552371MCP
get_claim(id=552371)