Generative AI video models are evolving rapidly, changing the landscape of content creation and presenting new challenges for newsrooms worldwide. According to insights penned by Storyful senior journalist Rachel Bidock, ambitious tech companies are racing to position themselves at the forefront of this new era, marketing their models as creative outlets for everything cinematic and absurd. However, malicious actors are actively exploiting these technologies, leading to a measurable surge in AI-generated user-created content circulating across social media platforms during breaking news events and viral moments. Specifically, Storyful has noted a 60 percent increase in video verification requests since the start of the US-Israel war with Iran. Despite the growing sophistication of leading models—such as OpenAI's Sora, Google's Veo 3.1, and ByteDance's Seedance 2.0—they all share fundamental weaknesses that serve as critical detection points. Using the iconic Will Smith eating spaghetti clip as a benchmark, analysts have evaluated the flaws of each iteration. Sora 2 displays static-filled spaghetti, misaligned audio, and physics errors where noodles unexpectedly fracture. Google's Veo 3.1 produces smooth features with cartoonish audio, alongside morphing noodles and physics inconsistencies. Meanwhile, ByteDance's Seedance 2.0 emerges as the most convincing model yet, though it still suffers from unrealistic noodle behavior, static-heavy teeth, and anatomical inaccuracies. Furthermore, Seedance-generated content has already sparked legal friction, with ByteDance receiving a cease-and-desist letter from Paramount and a takedown demand from Disney over intellectual property infringement concerns. As these AI videos become increasingly convincing and widespread, understanding their persistent flaws is essential for video desks and audience verification teams. Source: Storyful #AIVideo #Newsroom #Deepfakes #GenerativeAI #Storyful
También lo cubren: Storyful.
Ver más en News Today