Subscribe

Step-Video-T2V

Step-Video-T2V is a roughly 30B-parameter open (MIT) text-to-video model, the largest open video model when released.

StepFun's release preceded Wan 2.1 by 11 days; Wan's smaller 1.3B/14B sizes were easier to run.

Date
Friday 14 February 2025
Lab
StepFun
Kind
model
Access
open weights

The Hugging Face repo was created on February 14, 2025, and holds 29.3B parameters in safetensors under the MIT license.

Sources

  1. huggingface.co/stepfun-ai/stepvideo-t2v

This record was checked against its sources on 6 October 2026. How we check