Time Adapter for Kandinsky-5 I2V

Overview

This repository contains a lightweight TA-only temporal adapter checkpoint for Kandinsky-5 image-to-video generation.

Lower target FPS typically leads to slower perceived motion, while higher target FPS leads to faster perceived motion.

The released checkpoint keeps the base DiT weights unchanged and adds only the time-adapter parameters for global FPS conditioning and latent-time control.

Use this checkpoint with the `time_adapter` branch of kandinsky-5 and the config configs/k5_i2v_time_adapter_step5750_ta_only.yaml.

Examples

The gallery below shows TA-only examples for pose motion control and forest motion control at different target FPS values.

The config follows the standard `kandinsky-5` layout and expects model files under weights/model, weights/vae, weights/text_encoder, and weights/text_encoder2.

Prompt
Target FPS: 15. A close-up of a green-eyed girl in a dress stands in the city. The girl smiles and waves. Then she turns her back and walks away.
Prompt
Target FPS: 30. A close-up of a green-eyed girl in a dress stands in the city. The girl smiles and waves. Then she turns her back and walks away.
Prompt
Target FPS: 60. A close-up of a green-eyed girl in a dress stands in the city. The girl smiles and waves. Then she turns her back and walks away.
Prompt
Target FPS: 15. Beautiful views of the forest with fog. The wind dispels the fog and the sun peeks out over the forest. The sun's rays illuminate a beautiful green forest.
Prompt
Target FPS: 30. Beautiful views of the forest with fog. The wind dispels the fog and the sun peeks out over the forest. The sun's rays illuminate a beautiful green forest.
Prompt
Target FPS: 60. Beautiful views of the forest with fog. The wind dispels the fog and the sun peeks out over the forest. The sun's rays illuminate a beautiful green forest.
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for kandinskylab/Time-Adapter