Improve model card: Add pipeline tag, library, and update links and content
#2
by
nielsr
HF Staff
- opened
This PR significantly enhances the model card by:
- Adding
pipeline_tag: text-to-videoto improve discoverability on the Hugging Face Hub, as Kandinsky 5.0 is a family of foundation models for video (and image) generation, with a focus on text-to-video. - Adding
library_name: diffusers, as the model is explicitly mentioned to be compatible with the Diffusers library, enabling the automated "how to use" widget. - Updating the Project Page link to
https://kandinskylab.ai/. - Updating the Technical Report link to the official arXiv paper:
https://arxiv.org/abs/2511.14993. - Updating the GitHub repository link to
https://github.com/kandinskylab/kandinsky-5. - Replacing the model card content with the more comprehensive and up-to-date README from the official GitHub repository, which includes details on additional model variants (Video Pro, Image Lite, Image Editing), expanded quickstart/inference sections, and an acknowledgements section. This provides a much more complete overview of the Kandinsky 5.0 family.
- Updating the initial logo to support dark/light mode as seen in the GitHub README.
- Updating the citation to reflect the full author list and correct arXiv link.
Please review and merge if these changes are appropriate.