On the afternoon of April 27, the Qwen app launched a gray-scale test of Alibaba's video model HappyHorse. Users can click the "HappyHorse" button at the bottom of the home page to experience it. With its outstanding narrative ability, audio-visual synchronization, and diverse style expression, during the internal testing period, creators generated a large number of short videos in TVB Hong Kong style, CCTV Three Kingdoms style, and old movie style, which were published on the Qwen app community. Users can also one-click create similar content using prompts.

1777294767681.jpg

HappyHorse has unique advantages in narrative-style videos. For instance, in terms of narrative ability, just a simple description can generate a multi-shot video with corresponding camera movements and transition effects; in terms of style expression, HappyHorse can accurately understand and reproduce various styles such as old Hong Kong movies and old films.

During the internal testing phase, creators have already "broken" HappyHorse on the Qwen app's AI creation community: for example, they used the style of the old CCTV version of "Romance of the Three Kingdoms" to generate "workplace nonsense" clips, where a strategist advised a general, "If you have a stomachache, eat more of the boss's drawings of big pancakes"; another netizen used the style of "Criminal Investigation Archive" to create a short film of a police officer interrogating a cat, sentencing "all the fish cans this month will be given to stray cats."

1777294788532.jpg

In the near future, the Qwen app will launch a new video feature called "Test Yourself," where completing a few simple quiz questions can determine your "main character" role in the short drama universe. Then, by uploading a photo, users can generate a short drama clip featuring themselves through HappyHorse 1.0. In addition, Qwen will launch the "Imagination Challenge" on April 28, inviting creators to join in the experience.

As Alibaba's latest released multimodal video generation model, Happy Horse 1.0 supports 15-second multi-shot storytelling, multi-aspect ratio adaptation, and 1080P super-resolution output. It excels in image quality, narrative ability, character performance, audio-visual synchronization, and style diversity, attracting high attention from the global AI community.