Overview

Unlock the future of video generation with LTX-2, the groundbreaking DiT-based audio-video foundation model. This official Python package provides everything you need for both advanced inference and powerful LoRA training, bringing synchronized audio and video, high fidelity, and production-ready outputs to your fingertips.

Key Features

• Comprehensive Inference: Leverage a suite of specialized pipelines for text-to-video, image-to-video, audio-to-video, lip-dubbing, and more.
• Advanced LoRA Training: Fine-tune and customize models with robust tools for LoRA, full fine-tuning, and IC-LoRA, enabling precise control over your video generations.
• High Fidelity & Synchronization: Experience synchronized audio and video with exceptional visual quality.
• Optimized Performance: Benefit from various optimization strategies, including FP8 quantization, attention optimizations, and gradient estimation for faster, more efficient inference.
• Extensible Architecture: Built with a monorepo structure for modularity, featuring ltx-core, ltx-pipelines, and ltx-trainer for flexible development.
• Production-Ready Outputs: Generate professional-grade video content with support for API access and multiple performance modes.
• Cutting-Edge Control: Utilize a wide array of LoRAs for fine-grained control over camera movements, details, and more.

Typical Use Cases

• Filmmakers & Content Creators: Generate dynamic video sequences from text prompts, create unique visual effects, and produce compelling marketing content.
• Researchers & Developers: Explore the forefront of audio-video generation, experiment with new training techniques, and build custom video AI applications.
• Prototyping & Experimentation: Rapidly iterate on video ideas with single-stage pipelines and optimized inference options.
• Interactive Media: Develop engaging experiences with capabilities like keyframe interpolation and retaking specific video segments.
• AI-Powered Dubbing & Revoicing: Create realistic lip-dubbed videos and rephrase existing dialogue with advanced IC-LoRA pipelines.