A China-based startup, Sand AI, has launched an overtly licensed video-generating AI mannequin that’s garnered reward from entrepreneurs like Microsoft Analysis Asia founding director Kai-Fu Lee. However Sand AI seems to be censoring photographs that may increase the ire of Chinese language regulators from the hosted model of the mannequin, in response to TechCrunch’s testing.
Earlier this week, Sand AI introduced Magi-1, a mannequin that generates movies by “autoregressively” predicting sequences of frames. The corporate claims the mannequin can generate high-quality, controllable footage that captures physics extra precisely than rival open fashions.
👀 You received’t consider that is AI-generated
🧠 You received’t consider it’s open-source
🎬 You received’t consider it’s FREEMagi-1 video mannequin simply humiliated industrial video instruments
Particulars and examples beneath: 👇 pic.twitter.com/zlXRecWeqH
— Farhan (@mhdfaran) April 21, 2025
Magi-1 is simply too impractical to run on most shopper {hardware}. It’s 24 billion parameters in measurement, and requires between 4 and eight Nvidia H100 GPUs to run. (Parameters are the interior variables fashions use to make predictions.) For a lot of customers — this reporter included — Sand AI’s platform is the one place they will check drive Magi-1.
The platform wants a “prompt” picture to kick off video era. Not all prompts are permissible, TechCrunch shortly found. Sand AI blocks picture uploads of Xi Jinping, Tiananmen Sq. and Tank Man, the Taiwanese flag, and insignias supporting Hong Kong liberation. Filtering seems to be taking place on the picture stage; renaming picture recordsdata didn’t skirt the blocking.
Sand AI isn’t the one Chinese language startup stopping uploads of politically delicate photographs to its video era software. Hailuo AI, Shanghai-based MiniMax’s generative media platform, blocks photographs of Xi Jinping as nicely. However Sand AI’s filtering seems to be notably aggressive; Hauiluo permits photographs of Tiananmen Sq..
As Wired defined in a bit from January, fashions in China are required to observe stringent data controls. A 2023 legislation forbids fashions from producing content material that “damages the unity of the country and social harmony” — that’s, counters the federal government’s historic and political narratives. To conform, Chinese language startups typically censor their fashions, both by means of prompt-level filters or fine-tuning.
Apparently, whereas Chinese language fashions have a tendency to dam political speech, they typically have fewer filters than their American counterparts for pornographic content material. 404 just lately reported that plenty of video turbines launched by Chinese language firms lack primary guardrails that forestall folks from producing nonconsensual nudity.
