Skip to content

Step-Video-TI2V

Control motion and camera work in image-to-video generation

Visit official site

What Step-Video-TI2V is

The model is described as an image-to-video model built on a 30B-parameter Step-Video-T2V base, able to generate up to 102 frames from text and image inputs. It is stated to balance dynamics and stability through controllable motion amplitude and camera movement, with weights and inference code released on GitHub.

What you can do with it

  • Set a motion score to move from stable frames to highly dynamic action
  • Choose fixed, pan, tilt, zoom, push-in, pull-out, rotation, tracking or orbit camera work
  • Generate landscape, portrait and square video from a single image
  • Produce high-resolution clips with smooth motion and anime-style effects

Application scenarios

  • Turn an anime character image and a text description into an animated video
  • Specify orbit, push-in or pull-out moves for a film-like clip used in video work or advertising
  • Generate effect scenes such as lightning and spirit beasts for a fantasy video

Getting started

  • Download the model weights and inference code from the GitHub page
  • Create a virtual environment and install the required Python packages
  • Prepare an image as the first frame and a text description to guide the content
  • Adjust the motion score and camera work parameters before generating

Pros & cons

✓ What we like

  • Motion amplitude and camera movement are both adjustable
  • Weights and inference code can be downloaded from GitHub

! What to watch out for

  • Video length is stated as up to 102 frames
  • Inputs need both an image for the first frame and a guiding text description

FAQ

What does the model generate?

The material states text and image inputs produce videos of up to 102 frames.

Where are the model files?

The material says weights and inference code can be downloaded from the GitHub page.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review