LTX-2-19b

LTX-2 19B is Lightricks' large scale open source video generation model, built around an asymmetric dual stream transformer with 14 billion parameters for video and 5 billion for audio, 19 billion total, coupled through bidirectional cross attention. It is the first open weights model to produce synchronized audio and video in a single pass, with native 4K output at up to 50 frames per second and clip lengths up to 20 seconds. The audio stack covers speech, foley, ambience, and emotional elements aligned to the visual content, with lip sync for character dialogue. LTX-2 is distributed with open weights, which supports research, experimentation, and custom deployment for teams that want full control over a high end video generation pipeline.

January 6, 2026

Lightricks

Text to Video (T2V)