Text2Video-Zero: Text-to-Video Zero Sample Generator Released by the Picsart AI Research Team

General Introduction

Text2Video-Zero is an official implementation of a zero-sample text-to-video generator for GitHub developed by the Picsart AI Research team.The project provides a new way to use text cues to generate videos with temporal consistency and correctly followed text cues. The team has also released a companion academic paper with associated demos and code. Through this project, Picsart AI Research hopes to contribute to the development of AI research and creative expression.

Text2Video-Zero Online Experience

Function List

Zero-sample video generation using text prompts
Provide sample code and runtime methods
Support for different stabilized diffusion base model loading
Provides a variety of hyper-parameter customization
Supports operation in low memory configurations
Provides different types of video control generation
Release of latest updates and code optimizations
Provide links to presentations and papers

Using Help

Clone the Github repository and follow the documentation to install the environment
Run the reasoning example to generate a video
Setting customized hyperparameters
Explore pose, edge, dreambooth and depth controlled video generation.
Refer to the changelog for the latest features and optimizations
Read the paper to get the technical and research background