Happy New Year
Lightricks Open Source

LTX-2: 4K AI Video with Synchronized Audio

19B Parameters · 4K 50FPS · 20 Seconds · Apache 2.0

The first complete open-source AI video foundation model. Generate cinematic 4K videos at 50FPS with perfectly synchronized audio, dialogue, and sound effects in a single pass.

Lightricks Open Source

Generate with LTX-2

Enter a rich prompt to create a synchronized video. Credits required.

HOT

Generate with LTX-2

Enter a rich prompt to create a synchronized video. Credits required.

Durations over 10s must use 1080p resolution.

Available credits0
-30Cost
0Available

Recent tasks

DiT Architecture

What is LTX-2?

LTX-2 is a groundbreaking open-source AI video model developed by Lightricks, featuring a 19-billion parameter Diffusion Transformer (DiT) architecture. It generates synchronized 4K video and audio in a single unified process.

With its asymmetric dual-stream design (14B video + 5B audio parameters), LTX-2 produces cinematic-quality content with natural dialogue, ambient sounds, and music perfectly aligned to on-screen action.

4K UHD
Resolution
50 FPS
Frame Rate
20 sec
Max Duration
Audio Sync
Native

LTX-2 Model

Open Source Video Foundation Model

Synchronized audio-video generation in single pass

Native 4K resolution at up to 50 frames per second

Apache 2.0 license - fully commercial use allowed

Multimodal inputs: text, image, video, keyframes

Why Choose LTX-2?

The most advanced open-source AI video model available

Synchronized Audio-Video

Generate video and audio simultaneously in one pass. Natural alignment of dialogue, ambient sounds, music, and Foley effects with on-screen action.

Native 4K Resolution

Support for native 4K (3840×2160) resolution at up to 50 frames per second, delivering cinematic-quality visuals for professional production.

Extended Duration

Generate video clips up to 20 seconds long while maintaining temporal coherence and consistent quality throughout the entire sequence.

Open Source Apache 2.0

Fully open-source with model weights, training code, and inference pipeline available. Customize, fine-tune, and deploy locally as needed.

Multimodal Input

Accept various inputs including text prompts, images, depth maps, video snippets, audio samples, and multi-keyframe conditioning for precise control.

Efficient Performance

Optimized for speed with NVFP4/NVFP8 support. Up to 3x faster and 60% VRAM reduction on RTX 50 Series GPUs. Run on consumer hardware.

Technical Specifications

Industry-leading video generation capabilities

Specification
LTX-2

Model Parameters

19 Billion (14B Video + 5B Audio)

Maximum Resolution

4K (2160p UHD)

Frame Rate

Up to 50 FPS

Maximum Duration

6-20 seconds

Audio Synchronization

Native

License

Apache 2.0

Use Cases

Perfect for filmmakers, advertisers, and content creators

Film & VFX

Create pre-visualizations, mood films, concept videos, and VFX sequences. Plan camera movements and explore visual directions before production.

  • Pre-visualization
  • Concept Development
  • VFX Previews

Advertising & Marketing

Produce high-quality promotional videos rapidly. Test multiple concepts, generate product visualizations, and create campaign content efficiently.

  • Product Videos
  • Brand Campaigns
  • A/B Testing

Social Media

Create engaging short-form content for TikTok, Reels, YouTube Shorts. Stand out with cinematic AI-generated videos with synchronized audio.

  • TikTok / Reels
  • YouTube Shorts
  • Story Content

Simple 3-Step Workflow

From idea to video in minutes

Describe or Upload

Enter a text prompt or upload an image/video to guide the generation.

AI Generates

LTX-2 creates synchronized 4K video with audio in a single pass.

Download & Share

Get your high-quality video ready for any platform or project.

19B
Parameters
4K
Resolution
50
FPS
20s
Max Duration

Frequently Asked Questions

LTX-2 is an open-source AI video foundation model developed by Lightricks. It features 19 billion parameters and generates synchronized 4K video and audio in a single pass using Diffusion Transformer (DiT) architecture.

LTX-2 supports native 4K resolution (3840×2160) at up to 50 frames per second, delivering cinematic-quality output suitable for professional production workflows.

LTX-2 can generate videos from 6 to 20 seconds in length, maintaining temporal coherence and consistent quality throughout the entire sequence.

Yes, LTX-2 is fully open-source under the Apache 2.0 license. Model weights, training code, and inference pipeline are available for download, customization, and commercial use.

Yes, videos generated with LTX-2 can be used commercially. The Apache 2.0 license allows full commercial usage rights for marketing, advertising, and content creation.

Yes, LTX-2 generates synchronized audio alongside video in a single pass. This includes dialogue, ambient sounds, music, and Foley effects naturally aligned with the visual content.

Ready to Create with LTX-2?

Experience the most advanced open-source AI video generation.