ali-vilab/i2vgen-xl

RESEARCH/NON-COMMERCIAL USE ONLY: High-Quality Image-to-Video Synthesis via Cascaded Diffusion Models

Capabilities

Reference ImagesSeed

Community model (estimated from hardware time)

Name	Type	Description	Default	Constraints
`image`*	string(uri)	Input image.	`—`	—
`prompt`*	string	Describe the input image.	`—`	—
`guidance_scale`	number	Scale for classifier-free guidance	`9`	min: 1, max: 20
`max_frames`	integer	Number of frames in the output	`16`	min: 2
`num_inference_steps`	integer	Number of denoising steps	`50`	min: 1, max: 500
`seed`	integer	Random seed. Leave blank to randomize the seed	`—`	—

imagerequiredstring

Input image.

promptrequiredstring

Describe the input image.

guidance_scalenumber

Scale for classifier-free guidance

Default: 9min: 1, max: 20

max_framesinteger

Number of frames in the output

Default: 16min: 2

num_inference_stepsinteger

Number of denoising steps

Default: 50min: 1, max: 500

seedinteger

Random seed. Leave blank to randomize the seed

Version: 5821a338d000Updated: 7/25/2026128.3K runs