Mora is an open-source, multi-agent framework designed for generalist video generation, aiming to replicate and extend the capabilities of models like OpenAI's Sora. It targets researchers and developers interested in advanced video synthesis, offering a collaborative approach with specialized visual agents to handle diverse tasks from text-to-video to simulating digital worlds.

How It Works

Mora employs a multi-agent system where distinct AI agents collaborate on different aspects of video generation. This modular design allows for specialization, potentially leading to higher quality outputs across a broad spectrum of video synthesis tasks. The framework is built to be extendable, encouraging community contributions and customization.

Quick Start & Requirements

Code release is pending. The project is associated with the paper "Mora: Enabling Generalist Video Generation via A Multi-Agent Framework."

Highlighted Details

• Achieves performance comparable to Sora in various video generation tasks.
• Supports text-to-video, image-to-video, video extension, video-to-video editing, video connection, and digital world simulation.
• Demonstrates video generation up to 12 seconds at 1024x576 resolution.
• Aims to replicate Sora's capabilities, including generating up to 80-second videos.

Maintenance & Community

The project is associated with researchers from Lehigh University. Updates are planned, with v2 paper and training code expected soon. Contact is available via email for suggestions.

Licensing & Compatibility

The repository is open-source, but the specific license is not detailed in the README. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

While aiming for Sora-level capabilities, Mora currently has a significant gap in resolution, object consistency, and motion smoothness compared to Sora. The code is not yet released.