Official announcements from Tencent or updates on their open-source platforms

Tencent Releases 13-Billion-Parameter HunyuanVideo, Largest Open-Source Text-to-Video AI Model

Dec 3, 2024, 08:42 AM

Tencent Holdings Ltd. has released HunyuanVideo, a 13-billion-parameter open-source text-to-video AI model, which is currently the largest of its kind. HunyuanVideo is designed to generate high-quality, realistic videos from text inputs and is reported to be competitive with leading closed-source models like Runway Gen-3. The model incorporates advanced features such as an MLLM Text Encoder and 3D VAE, enhancing its capabilities in video generation. HunyuanVideo supports various functions, including text-to-video generation, image-to-video transformation, avatar animation, and dynamic scene modeling, suitable for industry-level applications like advertising and creative video production. By releasing both code and weights, Tencent positions itself at the forefront of open-source AI development, reflecting China's growing influence in this field.