CascadeV

by ByteDance

About the model:

Video generation model implementing the Würschten architecture.

Model type:

Video

Model performance class:

Limited

Link to the model:

https://huggingface.co/ByteDance/CascadeV

Base models:

PixArt-Sigma-XL-2-1024-MS

End model:

CascadeV

End model license:

CreativeML Open RAIL++-M License

About the organisation:

ByteDance, the parent company of TikTok.

Link to the organisation:

https://www.bytedance.com/

Model release date:

September 2024

Availability

Base Model Data

Internal dataset, apparantly undisclosed.

End User Model Data

No data sources found.

Base Model Weights

Weights made available through HuggingFace.

https://huggingface.co/PixArt-alpha/PixArt-Sigma-XL-2-1024-MS

End User Model Weights

Weights made available on HuggingFace.

https://huggingface.co/ByteDance/CascadeV

Training Code

Training code published on GitHub.

https://github.com/bytedance/CascadeV

Documentation

Code Documentation

Training code sparsely documented.

Hardware Architecture

No hardware architecture information found.

Preprint

Preprint available on arXiv.

https://arxiv.org/pdf/2306.00637

Paper

Paper submitted to ICLR 2024 oral.

https://openreview.net/forum?id=gU58d5QeGv

Modelcard

Model card contains most required information, but lacks details on training data and fine-tuning.

https://huggingface.co/ByteDance/CascadeV

Datasheet

No datasheet found.

Access

Licenses

CreativeML Open RAIL++-M License. Not an OSI-approved license.

https://huggingface.co/ByteDance/CascadeV

Is this information not up to date?

Contribute here ->

Supported by the Centre for Language Studies and the Dutch Research Council. Website design & development © 2024 by BSTN. This version of the index generated 09 April 2026, website content last updated 11 March 2026.