Breaking

Lightricks Releases LTX-2, an Open-Weights Model for Synchronized Audio and Video Generation

October 6, 2026 at 08:21 EDT

Media Generation
Open Source
Foundation Models

Lightricks has released LTX-2, an open-weights audio-video foundation model designed to generate synchronized sound and visuals in a single model rather than adding audio after video creation. The release positions LTX-2 as a practical local and workflow-ready system for creators using ComfyUI, Diffusers and Lightricks' own tooling.

Oct. 23, 2025 · Lightricks

One model for synced video and audio

LTX-2 generates visuals and sound together, bringing 4K, up to 50 FPS video, multi-keyframe control and LoRA customization into an open audio-video model family.

19B

full LTX-2 checkpoint

native resolution target

FPS maximum cited

10s

continuous synced clips

Model capacity is weighted toward video

14B

video stream

audio stream

The asymmetric design dedicates more capacity to visual synthesis while keeping audio linked through cross-attention.

Reported inference gap

Wan 2.2 baseline

18x

LTX-2 evaluation

What creators gain

One workflow can combine prompts, images, keyframes, LoRAs and upscalers while keeping audio timing inside generation.

What still takes care

Users need matching checkpoints and workflow versions, and frame-rate settings can affect perceived sync.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.