Dropped from the LTX API in August 2026, so it is mainly for researchers running the open weights locally.

Kenning's take

What it does

LTX-2 is an open-source video model from Lightricks that generates video and matching sound together, at up to native 4K and 50 frames per second, in clips of around 10 seconds, on consumer graphics cards. It was announced in October 2025 and its full weights were released in January 2026. It has since been replaced by LTX-2.3 and LTX-2.5: the LTX API stopped serving LTX-2 on 15 August 2026, but the weights remain on Hugging Face.

  • Synchronised audio and video generation
  • Native 4K at up to 50 fps
  • About 10-second clips
  • Camera, depth and pose control
  • LoRA training support

Who it suits. Developers and researchers who want to run or study an earlier open video model locally.

Key facts

Category
AI models
Part of
LTX Studio
Made by
Lightricks
Released
23 October 2025
How to get it
open weights on Hugging Face (released 5 January 2026); no longer available in the LTX API since 15 August 2026
Works on
self-hosted, Hugging Face
In the UK
Yes
Free plan
open weights on Hugging Face

FAQ

When was LTX-2 released?

Lightricks announced LTX-2 on 23 October 2025 and released the full weights on 5 January 2026.

Can I still use LTX-2 in the API?

No. LTX-2 requests were moved to LTX-2.3 on 15 July 2026 and the model was removed from the API on 15 August 2026.

What replaced LTX-2?

LTX-2.3, released in March 2026, and LTX-2.5, released in August 2026.

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.