Alibaba makes Sora-style video models open source, as tech giant shows AI capabilities
Alibaba Group will open source four Sora-style video-generation models from its Tongyi Wanxiang series to promote its AI technologies among developers. The move by Alibaba Cloud, the company's cloud computing and AI arm, makes the Wan2.1 series models freely available to academics, researchers, and commercial institutions globally.
These models, trained on 1.3 billion or 14 billion parameters, can be downloaded from Alibaba's Model Scope and Hugging Face. This open-source approach, gaining traction due to companies like DeepSeek, allows developers to modify, share, and improve the models, potentially fostering a wave of new AI applications. The Wan2.1 series, already a leading video-generation model supporting text effects in both Chinese and English, aims to rival the performance of models from companies like OpenAI.
via SCMP Full Text Feed
Alibaba Group will open source four Sora-style video-generation models from its Tongyi Wanxiang series to promote its AI technologies among developers. The move by Alibaba Cloud, the company's cloud computing and AI arm, makes the Wan2.1 series models freely available to academics, researchers, and commercial institutions globally.
These models, trained on 1.3 billion or 14 billion parameters, can be downloaded from Alibaba's Model Scope and Hugging Face. This open-source approach, gaining traction due to companies like DeepSeek, allows developers to modify, share, and improve the models, potentially fostering a wave of new AI applications. The Wan2.1 series, already a leading video-generation model supporting text effects in both Chinese and English, aims to rival the performance of models from companies like OpenAI.
via SCMP Full Text Feed
