Tencent Releases Hunyuan Video: 130B Parameter SOTA Model for High-Def Video Generation

Dec 3, 2024, 05:17 AM

Tencent Holdings Ltd. has released Hunyuan Video, a new open-source video generation model, marking a significant advancement in AI-driven video content creation. The model, which boasts 130 billion parameters, is now available with its weights, allowing developers and researchers to utilize and build upon it. Hunyuan Video excels in generating high-definition, realistic videos suitable for commercial applications such as advertising and creative video production. It supports text-to-video generation, image-to-video conversion, and can produce sound for videos, among other capabilities. The model has been praised for its stunning outputs and is considered a state-of-the-art (SOTA) video generation model, surpassing existing models in several aspects. It uses a Sora-like DiT architecture and a new language model for text encoding, enhancing its semantic adherence.