\n<\/p>\n
A China-based startup, Sand AI<\/a>, has released an openly licensed video-generating AI model that\u2019s garnered praise from entrepreneurs like Microsoft Research Asia founding director Kai-Fu Lee. But Sand AI appears to be censoring images that might raise the ire of Chinese regulators from the hosted version of the model, according to TechCrunch\u2019s testing.<\/p>\n Earlier this week, Sand AI announced Magi-1<\/a>, a model that generates videos by\u00a0\u201cautoregressively\u201d predicting sequences of frames. The company claims the model can generate high-quality, controllable footage that captures physics more accurately than rival open models.<\/p>\n \ud83d\udc40 You won\u2019t believe this is AI-generated Magi-1 video model just humiliated commercial video tools<\/p>\n Details and examples below: \ud83d\udc47 pic.twitter.com\/zlXRecWeqH<\/a><\/p>\n \u2014 Farhan (@mhdfaran) April 21, 2025<\/a><\/p>\n<\/blockquote>\n Magi-1 is too impractical to run on most consumer hardware. It\u2019s 24 billion parameters in size, and requires between four and eight Nvidia H100 GPUs to run. (Parameters are the internal variables models use to make predictions.) For many users \u2014 this reporter included \u2014 Sand AI\u2019s platform is the only place they can test drive Magi-1.<\/p>\n The platform needs a \u201cprompt\u201d image to kick off video generation. Not all prompts are permissible, TechCrunch quickly discovered. Sand AI blocks image uploads of Xi Jinping, Tiananmen Square and Tank Man, the Taiwanese flag, and insignias supporting Hong Kong liberation. Filtering appears to be happening at the image level; renaming image files didn\u2019t skirt the blocking.<\/p>\n Sand AI isn\u2019t the only Chinese startup preventing uploads of politically sensitive images to its video generation tool. Hailuo AI, Shanghai-based MiniMax\u2019s generative media platform, blocks photos of Xi Jinping as well. But Sand AI\u2019s filtering appears to be particularly aggressive; Hauiluo allows images of Tiananmen Square.<\/p>\n As Wired explained in a piece from January, models in China are required to follow stringent information controls. A 2023 law forbids models from generating content that \u201cdamages the unity of the country and social harmony\u201d \u2014 that is, counters the government\u2019s historical and political narratives. To comply, Chinese startups often censor their models, either through prompt-level filters or fine-tuning. <\/p>\n Interestingly, while Chinese models tend to block political speech, they often have fewer filters than their American counterparts for pornographic content. 404 recently reported<\/a> that a number of video generators released by Chinese companies lack\u00a0basic guardrails that prevent people from generating nonconsensual nudity.<\/p>\n<\/div>\n\n
\ud83e\udde0 You won\u2019t believe it\u2019s open-source
\ud83c\udfac You won\u2019t believe it\u2019s FREE<\/p>\n