Alibaba has launched a limited beta for its video generation model, HappyHorse 1.0, powered by a native multimodal architecture with integrated audio-visual generation for ads, e-commerce, short dramas, and social media content.

In a hands-on test conducted by National Business Daily, an AI-generated nostalgic schoolgirl image was used as the first frame, along with a prompt. The materials were fed into both Alibaba’s Wanxiang and HappyHorse models. In about three minutes, an 8–10 second video was successfully generated.

The test suggests that HappyHorse can deliver practical results under clear input conditions. According to Li Ming, a technical partner at Max Engine, outputs are generally usable when the first frame is clean and prompts are precise. Character transitions in comic-style scenes were also relatively smooth.

Li Ming said that creators are primarily focused on cost, speed, and consistency. For many small and mid-sized teams producing large volumes of content daily, current AI video generation costs remain a major constraint.