OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Open Genie

Open Genie is an unofficial, MIT-licensed Python and PyTorch implementation of Genie, a model for learning generative interactive environments from unlabeled videos. It provides LightningCLI training workflows for a VideoTokenizer and for the LatentAction and Dynamics components.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Open Genie — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use the repository when you want to work directly with Genie model code and run its documented training stages in a local Python environment.

Who it is for

  • Researchers exploring generative interactive environments learned from video.
  • Python and PyTorch developers who want to train or customize Genie model components.

Documented uses

  • Train a VideoTokenizer that converts video into discrete tokens.
  • Train LatentAction and Dynamics components after training the VideoTokenizer.
  • Experiment with configuration-driven training of a Genie-style model.
Before you choose

Requirements & limits

The project is a public GitHub repository with an MIT license.

  • The project is explicitly an unofficial implementation of Genie.
  • The README specifies Python 3.11+ and PyTorch 2.0+ and uses a LightningCLI interface.
  • Training LatentAction and Dynamics depends on a fully trained VideoTokenizer.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Train the VideoTokenizer

Start training the Genie VideoTokenizer with a repository configuration file.

Before you start

  • Python 3.11 or later and PyTorch 2.0 or later.
  • Install dependencies with pip install -r requirements.txt.
  • A configuration file; example configuration files are provided in the config folder.

Steps

  1. Choose or prepare a configuration file for the VideoTokenizer.

  2. Run the command below.

    Documented command
    python tokenizer.py train -config

Expected outcome

The LightningCLI command starts the VideoTokenizer training workflow using the selected configuration.

Guide limitations

  • This command covers VideoTokenizer training only; the README does not provide a dataset-preparation procedure.
  • Training LatentAction and Dynamics is a separate stage and relies on a fully trained VideoTokenizer.
Documentation 1 ↗
Source-based practical guide

Train LatentAction and Dynamics

Use the repository’s CLI to train the LatentAction and Dynamics components after preparing the VideoTokenizer stage.

Before you start

  • Python 3.11 or later and PyTorch 2.0 or later.
  • Install dependencies with pip install -r requirements.txt.
  • A fully trained VideoTokenizer and a configuration file; example configuration files are provided in the config folder.

Steps

  1. Prepare a configuration file for the Genie training command.

  2. Run the command below.

    Documented command
    python genie.py train -config

Expected outcome

The command starts training for the LatentAction and Dynamics models through LightningCLI.

Guide limitations

  • The README states that this stage relies on a fully trained VideoTokenizer.
  • The command does not document video-data preparation or a post-training inference procedure.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
myscience/open-genie · Official homepage title · retained in its source language
How we review sourcesSuggest a correction