Press releases or reports from major advertising firms or Tencent

Tencent Releases 13B-Parameter HunyuanVideo Model, Open-Source Alternative to Runway Gen-3

Dec 3, 2024, 07:25 AM

Tencent has announced the release of HunyuanVideo, a novel open-source video generation model with 13 billion parameters, making it the largest of its kind to date. Capable of producing high-quality, realistic videos from text prompts, the model supports applications such as advertising and creative video generation in industrial-grade commercial scenarios. HunyuanVideo utilizes advanced features including a new language model for text encoding, an MLLM Text Encoder, 3D VAE, and is based on a DiT (Denoising Diffusion Transformer) architecture. It is considered competitive with leading closed-source models such as Runway Gen-3 and Luma 1.6, and offers functionalities like animating talking avatars and dynamic scene modeling. The release includes both the code and model weights, allowing for wider adoption and experimentation within the AI community.