Zhipu AI·open·CogVideoX License
CogVideoX-5B
Open CogVideoX 5B expert-transformer text-to-video checkpoint from Zhipu AI / Tsinghua (QingYing lineage). Generates ~6s 720×480 clips at 8 fps; later CogVideoX1.5 variants extend resolution/duration. Self-host via Diffusers/GitHub.
- Max res 720 × 480
- Max duration 6s
- Price —
- Speed —
Specs
- License
- CogVideoX License
- Open weights
- Yes
- Max resolution
- 720 × 480
- Max duration
- 6s
- Generation
- —
- Release
- Aug 27, 2024
Benchmarks
Artificial Analysis Video Arena Elo
Blind pairwise preference Elo for text-to-video models from the Artificial Analysis Video Arena (with-audio leaderboard). Scores are only comparable within the same Arena protocol and voter pool. (source)
—
unreported