Skip to content

Open-Sora Plan v1.2

An open-source text-to-video generation model

Visit official site

What open-sora-plan-v1-2 is

The model is an open-source video generation system focused on turning text into video clips.

What you can do with it

  • Create video content from text prompts
  • Train on a prepared video dataset
  • Evaluate output with quality metrics

Key features

  • A 3D full attention architecture
  • An optimized CausalVideoVAE structure
  • EMA weight finetuning during training

Getting started

  • Visit the GitHub page for model details
  • Install the required libraries and tools
  • Set up the training environment and dataset

Pros & cons

✓ What we like

  • 3D full attention captures joint spatial-temporal features
  • Code, data and models are released openly

! What to watch out for

  • Training requires setting up an environment and dataset
  • Evaluation uses PSNR, SSIM and LPIPS metrics

FAQ

What architecture is used?

The material names a 3D full attention architecture and a 2+1D design.

What dataset is named?

The Kinetic400 video dataset is listed for training.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review