kandinskylab

/

Kandinsky-5.0-T2V-Pro-pretrain-10s

like 2

Follow

Kandinsky Lab 202

This PR significantly improves the model card for the Kandinsky 5.0 family of models by:

• Adding pipeline_tag: text-to-video to the metadata, which improves discoverability on the Hugging Face Hub.
• Specifying library_name: diffusers to enable the automated "how to use" widget, as the kandinsky library is compatible with the diffusers ecosystem.
• Including the project logo, a comprehensive navigation bar, and the main title from the GitHub README for better context and branding.
• Linking directly to the official paper: Kandinsky 5.0: A Family of Foundation Models for Image and Video Generation.
• Adding links to the project page (https://kandinskylab.ai/) and the GitHub repository (https://github.com/kandinskylab/kandinsky-5).
• Integrating an overview video from the GitHub README to visually showcase model capabilities.
• Providing detailed descriptions of the "Kandinsky 5.0 Video Pro" and "Kandinsky 5.0 Video Lite" model families, including their specific features, model tables, and examples.
• Adding a practical "How to Use" section with installation instructions and a Python code snippet for text-to-video inference, directly sourced from the GitHub README.
• Including the academic citation for the paper and the acknowledgements section from the original repository.

Please review and merge this PR to make this model more accessible and informative for the community.