LTX releases LTX-2.5, an AI video model generating clips in 6.8 seconds

LTX, the generative media company that spun out of Israeli app maker Lightricks, has dropped LTX-2.5, the latest iteration of its open-weights video and “world” model. The headline number: it can generate video clips in 6.8 seconds, a speed that puts it in rarefied air among open-source video generation tools.

The model arrives with native integration into ComfyUI, the node-based workflow environment that has quietly become the standard prototyping tool for anyone serious about open generative media.

What LTX-2.5 actually delivers

The model is available immediately in three forms: as open weights on Hugging Face, inside ComfyUI through a day-one launch partnership, and via the LTX API for teams that prefer managed infrastructure.

The licensing structure is designed to be startup-friendly. Organizations pulling in less than $10 million in annual recurring revenue can use the model for free. Companies above that threshold need to negotiate a commercial license.

LTX says its model family has now surpassed 33 million downloads, which the company claims makes it the most-used open “world” model line on the market.

The road from Lightricks to open-weights video

LTX’s lineage traces back to Lightricks, the company behind consumer photo and video editing apps like Facetune. The spinout first released a text-to-video model in 2024, establishing itself as an early mover in the open-source video generation space.

The model has evolved steadily since then. LTX-2 landed in October 2025, followed by an interim version, LTX-2.3, in March 2026. Each release has layered on new capabilities, including synchronized high-resolution audio-video output and improvements to local-run efficiency, meaning users can generate video on their own hardware without shipping data to a cloud endpoint.

The 6.8-second generation time positions LTX-2.5 as one of the fastest options in a field where competing models can take significantly longer to produce comparable output.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.