Back to list
modelVideo generation modelopen sourcemodel
DiT-based audio and video basic model: LTX-2 has been open source, 19B, directly outputs picture + narration + live audio video
LTX-2 is an audio and video generation model based on DiT. It is open source and supports direct output of pictures, narration and live sound effects. This model is about 18 times faster than Wan 2.2-14B on NVIDIA H100. It is suitable for quickly generating short videos and advertisements, but it may be confusing when multiple people dialogue.
23 views0 stars3/5/2026
LTX-2 is an audio and video generation model based on DiT. It is open source and supports direct output of pictures, narration and live sound effects. This model is about 18 times faster than Wan 2.2-14B on NVIDIA H100. It is suitable for quickly generating short videos and advertisements, but it may be confusing when multiple people dialogue.