Open-Sora-Plan: video generator — what it does and what you need to run it

Open-Sora-Plan is an open-source video generator based on text descriptions, created by the PKU-YuanGroup team. The project aims to reproduce OpenAI's Sora model through community efforts, providing researchers and developers with an accessible tool for experimenting with video generation.

What it does

The authors describe the project as an attempt to reproduce a text-to-video (T2V) model. Version 1.3.0 claims support for WFVAE components, a prompt refiner, data filtering strategy, sparse attention, and a bucket-based training strategy. Specific quality metrics or the maximum length of generated video are not specified in the fact sheet.

What you need to run it

The code is written in Python and distributed under the MIT license. The repository is available at github.com/PKU-YuanGroup/Open-Sora-Plan. According to the developers, working with 93x480p resolution requires 24 GB of VRAM. Running on your own hardware is possible; a cloud version is not mentioned.

Who it is for

The tool is aimed at researchers and developers who want to study the architecture of video generators or fine-tune the model for their own tasks. The open source code and MIT license allow free modification of the project. For end users expecting a ready-made application with an interface, the project may be too low-level — the authors do not indicate the presence of a graphical shell or web interface.

The project is actively developed: the latest code change is dated March 2026, and the latest release is version 1.5.0. This indicates that the tool continues to receive updates, although the authors do not disclose details about the innovations in the fact sheet.