AI video generation on macOS is addressed by this native application, which leverages the LTX-2 model and Apple's MLX framework for optimized performance on Apple Silicon. It targets Mac users seeking a streamlined, powerful tool to create videos from text or images, complete with synchronized audio, voiceover, and music capabilities.

How It Works

This project utilizes SwiftUI for a native macOS interface and Apple's MLX framework for efficient, on-device machine learning computations tailored for M-series chips. Video generation follows a two-stage pipeline: initial generation at half resolution, followed by upsampling and refinement to the target resolution. This MLX-based approach aims to maximize performance and resource efficiency on Apple Silicon hardware.

Quick Start & Requirements

• Installation: Download the latest release from the Releases page. Upon first launch, configure the Python installation path in Preferences and validate the setup. Missing Python packages (e.g., mlx, mlx-video-with-audio) can be installed via a one-click option within the app.
• Requirements: macOS 14.0 or later, an Apple Silicon Mac (M1/M2/M3/M4 series), 32GB RAM minimum (64GB+ recommended), Python 3.10+ installed, and approximately 20-42GB of disk space for model weights.
• First Generation: The selected LTX-2 model (~22GB to ~48GB) is downloaded from Hugging Face on the first generation attempt, which may take 15-30 minutes.

Highlighted Details

• Native macOS application built with SwiftUI for a seamless user experience.
• Optimized for Apple Silicon using the MLX framework, enabling efficient on-device AI video generation.
• Features integrated synchronized audio generation with video, with options for local (MLX-Audio) or cloud-based (ElevenLabs) voiceover and AI-generated background music.
• Includes an optional "Gemma Prompt Enhancement" feature that rewrites text prompts for more detailed and optimized video generation.
• Provides a generation queue, history management, and customizable generation parameters for fine-tuning output.

Maintenance & Community

The provided README does not specify community channels (e.g., Discord, Slack) or highlight notable contributors, sponsorships, or partnerships.

Licensing & Compatibility

The project is released under the MIT License. This license generally permits commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The application has substantial hardware requirements, including a minimum of 32GB RAM (64GB+ recommended) and significant disk space for model downloads. Some LTX LoRA workflows may not be directly compatible with the MLX-converted audio-video path without additional conversion tooling. The project notes that LTX-2.3 models are currently in Beta.