Zhipu AI to Open Source CogVideoX Video Generation Model

TapTechNews August 6th news, Zhipu AI announced today that it will open source the video generation model - CogVideoX which is homologous to Qingying.

It is introduced that the CogVideoX open source model contains multiple models of different sizes. Currently, CogVideoX-2B will be open sourced. Its inference under FP-16 precision requires 18GB of video memory, and fine-tuning requires 40GB of video memory, which means that a single 4090 graphics card can perform inference, and a single A6000 graphics card can complete fine-tuning.

CogVideoX-2B has a prompt token limit of 226, the video length is 6 seconds, the frame rate is 8 frames per second, and the video resolution is 720*480.

The official said that models with stronger performance and larger number of parameters are on the way. Please pay attention and look forward to it.

TapTechNews attached relevant links:

Code repository: https://github.com/THUDM/CogVideo

Model download: https://huggingface.co/THUDM/CogVideoX-2b

Technical report: https://github.com/THUDM/CogVideo/blob/main/resources/CogVideoX.pdf