Skip to content

Video-CCAM

Video language models with causal attention masks

Visit official site

What video-ccam is

The project is described as a family of flexible video language models. It targets understanding of both short and long videos through causal cross attention masks.

What you can do with it

  • Work on short video analysis tasks
  • Examine longer clips with the same models
  • Follow a tutorial notebook for inference

Key details

  • Benchmarks named include MVBench and MLVU
  • Source code was rewritten for simpler setup
  • Inference uses Huggingface transformers on NVIDIA GPUs

Pros & cons

✓ What we like

  • Source code was rewritten to simplify setup
  • Tutorials and examples are provided

! What to watch out for

  • A README and notebook guide the setup
  • An NVIDIA GPU is assumed for inference

FAQ

Who developed it?

A Tencent QQ multimedia research team is named.

What technique is used?

Causal cross attention masks.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review