The HappyHorse-1.0 video generation model, crafted by Alibaba’s ATH Innovation Business Division, is now undergoing testing on the Arena.ai platform and is in the last stages of refinement prior to its formal release. An updated version is anticipated to be launched in two weeks, with the API scheduled to go live on April 30. Additionally, early access has been provided to the community. Last week, the model took part in a blind test on a well-known AI platform under an undisclosed identity, outperforming ByteDance’s Seedance 2.0 and Kuaishou’s Kling 3.0 in the text-to-video and image-to-video (audio-free) categories, claiming the first-place spot. HappyHorse-1.0 utilizes a 40-layer single-stream self-attention Transformer architecture boasting 15 billion parameters, enabling native synchronized audio and video generation, along with native lip-sync capabilities across seven languages.