Short answer: Some video models can use images as visual references for the first frame, last frame, subject, or scene direction.
Steps
Choose a video model that supports image references.
Upload clear reference images.
Write a prompt explaining the motion and how the reference should be used.
Choose supported duration and resolution settings.
Start the generation and review the result.
Notes
Seedance models support reference workflows in Typeshutter.
A reference image controls visual direction; the prompt still needs to describe motion.
[Screenshot placeholder: video reference upload area.]
Troubleshooting
If the model ignores the reference, simplify the prompt and make the reference subject clearer.
If upload fails, try PNG or JPEG and reduce file size.