AI Deck

LTX-2 ─ Lightricks' Open-Weight Video Model That Generates Audio and Video Together

A video generation model developed by Lightricks (Israel). It supports both Text-to-Video and Image-to-Video, and its defining feature is unified video-plus-audio generation in a single pass, with dialogue and ambient sound synchronized to the motion on screen. It handles native 4K resolution at up to 50 frames per second. Following its 2025 announcement, January 2026 brought the release of the full package — model weights, training code, and inference pipeline — under a custom community license (the LTX-2 Community License), making local execution and integration into custom pipelines possible. There’s also a browser-based Playground to try it out, plus LTX Studio, a production environment covering everything from script to storyboard to editing. Note that March 2026 saw the release of the successor, LTX-2.3 (a 22B DiT, native vertical 1080×1920 support, a revamped VAE for finer detail, and improved audio quality from a new vocoder) — if you’re starting now, that’s the version to build on.

Key Features

  • Unified video and audio generation: Dialogue, sound effects, and ambient audio come out in a single generation pass, synchronized to on-screen motion and lip movement. No need to render the video first and add sound with a separate tool
  • Native 4K, up to 50fps: Native 4K resolution rather than upscaling, with smooth output at up to 50 frames per second
  • Both Text-to-Video and Image-to-Video: Beyond generating from text prompts, you can take an image you already have and animate it
  • Open weights: Model weights, training code, and inference pipeline are all published under a custom community license. Run it on a local GPU or wire it into a custom pipeline such as ComfyUI
  • Production workflow integration: LTX Studio takes you from script to storyboard, shot generation, and editing in one place. The browser Playground offers a lower-friction way to experiment

Pricing

LTX-2 is an open-weight model — you can pull the model itself from the public repository and run it locally (check the license terms). LTX Studio, the hosted environment, offers the following plans.

PlanMonthly priceKey features
Free$0Free tier for trying the basics
Lite$15For personal use, with a monthly compute allowance
Standard$35Commercial use permitted, larger compute allowance
Pro$125Large compute allowance, advanced features
EnterpriseContact salesCustom plans for larger teams

Pricing reflects information as of August 21, 2026. For details on generation volume and features included in each plan, check the official site.

Pros and Cons

Pros

  • Generates video with audio in a single pass, eliminating the separate sound-design step
  • Native 4K at 50fps — a high bar for output specs among video generation models
  • Weights, training code, and inference pipeline are all public, enabling local execution and integration into your own pipeline
  • Easy to try from the browser via Playground, while LTX Studio scales up into a full production workflow

⚠️ Cons

  • Running 4K at high frame rates locally demands substantial GPU power
  • The custom community license means you’ll need to verify commercial-use eligibility and conditions against your own situation
  • Serious cloud usage means committing to a paid LTX Studio plan sized to your generation volume
  • Compared to top-tier closed models, there are cases where consistency across complex scenes and instruction-following fall short

Comparison with Similar Services

CriteriaLTX-2SoraVeoRunway
ProviderLightricksOpenAIGoogleRunway
Model availabilityOpen weightsClosedClosedClosed
Simultaneous audio generationYesYesYesPartial
Local executionPossibleNoNoNo
Distinguishing trait4K, 50fps, custom pipeline integrationStrong at narrative generationGoogle ecosystem integrationIntegration with editing toolset

Who It’s For

  • Video creators who want to generate footage complete with dialogue and ambient sound from a single model
  • Developers and researchers looking to embed video generation into a local environment or in-house pipeline
  • Teams seeking a video generation model whose weights they control, rather than depending on a closed API
  • Solo creators who want storyboard through editing handled in one environment

Summary

LTX-2 stands out on three fronts: unified audio-video generation, native 4K, and open weights. In a video generation landscape dominated by closed commercial models, it publishes everything from weights to training code while still offering low-friction entry points through Playground and LTX Studio. Start with Playground to gauge generation quality, then decide between local execution and a paid LTX Studio plan when you’re ready to commit. This space turns over fast, so when you actually sit down with it, check whether the successor LTX-2.3 is available to you as well.

← Blog