LTX
LightricksLTX is Lightricks' family of diffusion-transformer video models. The current LTX-2 line, up to the 22B LTX-2.5, generates synchronized video and audio in one model from text, image, video or audio input, with native output at 1024 by 1536 and a 4K upscaling path; the earlier LTX-Video models were built for speed. Lightricks also publishes a trainer for LoRA and full fine-tuning.
Openness
3 high confidence- weights
- open(the LTX-2, 2.3 and 2.5 checkpoints and the LTX-Video line download from Hugging Face
- data
- closed(the cards say only that the models were trained on a large video dataset
- code
- open(the ltx-trainer package publishes LoRA, IC-LoRA and full fine-tuning
- license
- LTX-2.x-Community-License(LTX-2.5: commercial use free for entities under USD 10M annual revenue, a paid license above it, with use-based restrictions
- LTX-2 and 2.3 carry the LTX-2 Community License on the same terms)+LTXV-Open-Weights-License-0.X(LTX-Video 0.9.6 to 0.9.8, with the same USD 10M threshold)+Open-RAIL-M(LTX-Video 0.9.1 and 0.9.5
- use-based restrictions only)
LTX is free for commercial use by organizations under USD 10M in annual revenue, with a paid license above that. Lightricks publishes a fine-tuning trainer but not the pretraining code or data. LTX-2.x-Community-License and LTXV-Open-Weights-License-0.X allows commercial use only within a bound, so the license tier is use_bounded.
- https://huggingface.co/api/models?author=Lightricks&sort=downloads&direction=-1&limit=100 recorded 2026-09-26
Hugging Face listing of the Lightricks organization: the LTX-2, 2.3, 2.5 and LTX-Video checkpoints are public.
- https://huggingface.co/Lightricks/LTX-Video/raw/main/ltx-video-2b-v0.9.5.license.txt recorded 2026-09-26
LTX-Video 2B 0.9.5 license body: "Open RAIL-M License dated March 5, 2025".
- https://huggingface.co/Lightricks/LTX-Video/raw/main/LTX-Video-Open-Weights-License-0.X.txt recorded 2026-09-26
LTXV Open Weights License 0.X: "entities with annual revenues of at least $10,000,000 (the "Commercial Entities") are eligible to obtain a paid commercial use license".
- https://huggingface.co/Lightricks/LTX-Video/raw/main/README.md recorded 2026-09-26
LTX-Video card: "Trained on a large-scale dataset of diverse videos"; no dataset is released.
- https://raw.githubusercontent.com/Lightricks/LTX-2/main/LICENSE-2 recorded 2026-09-26
"LTX-2 Community License Agreement / License date: January 5, 2026", with the same USD 10,000,000 revenue clause.
- https://raw.githubusercontent.com/Lightricks/LTX-2/main/LICENSE-2_x recorded 2026-09-26
License body "LTX-2.x Community License Agreement / License date: August 11, 2026": "Entities with annual revenues of at least $10,000,000 (the "Commercial Entities") are required to obtain a paid license".
- https://raw.githubusercontent.com/Lightricks/LTX-2/main/README.md recorded 2026-09-26
README: "ltx-trainer - Training and fine-tuning tools for LoRA, full fine-tuning, and IC-LoRA".
Adoption
4 high confidenceAdoption is measured as Hugging Face downloads across the main LTX-2 and LTX-Video checkpoints, led by LTX-2.5 and LTX-2.3. The LoRA and upscaler repositories are left out.
- https://huggingface.co/api/models?author=Lightricks&sort=downloads&direction=-1&limit=100 recorded 2026-09-26
Organization listing sorted by downloads: LTX-2.5 1604804, LTX-2.3 1098602, LTX-Video 773634, LTX-2.3-fp8 548583 in the trailing 30 days.
Capability
3 medium confidenceLTX-2.5 places in the upper half of the video arena, ahead of Veo 3.1 on the silent board and second only to MiniMax H3 among downloadable models. MiniMax H3 sits among the three leaders overall, well above it.
- https://artificialanalysis.ai/video/leaderboard/text-to-video recorded 2026-09-26
Arena FAQ: "MiniMax H3 currently leads among open weights Text to Video models with audio ... followed by LTX-2.5 Fast (Elo 1055)"; without-audio leaderboard entry "LTX-2.5 Fast" (rank 16, Elo 1217.6).
Verified 2026-09-26