LTX-2.3: Previous Generation AI Video Model | LTX

Lightricks has updated its model lineup, designating LTX-2.3 as the previous-generation release while LTX-2.5 moves into the current position. The transition is a routine generational handoff rather than a deprecation - LTX-2.3 remains fully supported, and its open weights are still publicly available for developers and researchers who have integrated it into their workflows.

LTX-2.3 launched with a set of capabilities that distinguished it from many contemporaries in the open-weights video generation space. Its support for synchronized audio meant generated video could be paired with sound in a temporally coherent way, a feature that has practical value for content creators and product developers alike. Native portrait video support - optimized for vertical aspect ratios common on mobile platforms - was another practical inclusion that reflected real-world usage patterns rather than just benchmark-friendly outputs.

The shift to LTX-2.5 as the current model suggests Lightricks has made meaningful improvements across one or more of those dimensions, though the specifics of what changed between versions are best evaluated through direct comparison. For developers, the open-weights nature of both models means migration is a matter of choice rather than necessity - there is no forced sunset on the older release.

For anyone following the open-source video generation landscape, the LTX line is worth tracking. Lightricks occupies a relatively focused position in this space, building models with production and creator use cases in mind rather than purely academic benchmarks. With LTX-2.3 still available and LTX-2.5 now the active development target, users have the option to evaluate both and move on their own timeline.