LTX-2.3

LTX's previous-generation 22B video-and-audio model, still among the most downloaded video models. Free below US$10M revenue.

In short. LTX-2.3 is the March 2026 version of LTX's open video model. It makes video with matching sound from text, an image or audio, and it can make vertical 1080×1920 video. LTX-2.5 has replaced it, but it is still widely downloaded. It is free for organizations earning under US$10 million a year.

Summary

LTX released LTX-2.3 in early March 2026, with a blog post that calls it a major upgrade to the model architecture. LTX dates its press release 5 March, which this page uses as the release date; the Hugging Face repository was created on 4 March, and the blog post carries an earlier date of 23 February. It is a 22-billion-parameter diffusion transformer that generates video and synchronized audio in one model, and it takes text, images, audio or video as input. LTX lists four changes from LTX-2: a rebuilt VAE for sharper textures, hair and edges; a text connector four times larger for closer prompt following; less frozen output and fewer slow pans in image-to-video; and cleaner audio from filtered training data and a new vocoder. It is also the first LTX model with native vertical video, up to 1080×1920, trained on portrait data.

The Hugging Face repository holds the dev model, which is the one to fine-tune, and distilled checkpoints that need 8 steps and no guidance. It also has a distilled LoRA, two spatial upscalers and a temporal upscaler. Each checkpoint is one 46 GB BF16 file that bundles the transformer, the video and audio VAEs and the text projection, and the Gemma 3 text encoder is a separate download under Google's license. Lightricks publishes FP8 versions of about 29.5 GB in another repository, plus many task-specific IC-LoRAs for LTX-2.3. The model cards give no VRAM requirement. LTX Desktop, a free open-source video editor, launched alongside it.

LTX-2.5 replaced it on 11 August 2026 with multi-shot generation, a diffusion video decoder and a Gemma 4 text encoder, and its weights use the newer LTX-2.x Community License. The LTX-2.3 checkpoints stay under the LTX-2 Community License Agreement of 5 January 2026: free for organizations with annual revenue under US$10 million, counted with affiliates, and a paid agreement above that, plus a list of banned uses. Some third-party articles call LTX-2.3 Apache 2.0, which does not match LTX's own files. In the 30 days to 11 November 2026, Hugging Face counted about 884,000 downloads of the BF16 repository and 735,000 of the FP8 one, and those figures may include automated pulls.

Good for

  • Local video with synchronized sound if you already have workflows, LoRAs or IC-LoRAs built on LTX-2.3
  • Vertical 9:16 video up to 1080×1920 for mobile formats
  • Fine-tuning the dev model, which LTX says can take under an hour for a LoRA in many settings
  • Organizations under US$10 million in annual revenue

Watch out for

  • LTX-2.5 replaced it in August 2026, and the two generations' checkpoints are not interchangeable
  • Organizations at US$10 million of revenue or more, counted with affiliates, need a paid license, and the license bans a list of uses
  • The cards give no VRAM figure, and the checkpoints are 46 GB in BF16 and 29.5 GB in FP8 before the separate text encoder
  • Speed and quality claims come from LTX, and the card says audio without speech may be of lower quality

Performance claims come from the developer and have not been independently verified.

How much memory does it need?

Memory figures for each variant of LTX-2.3. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.

Variants Parameters Download size Memory needed Fits on
Dev and Distilled (BF16) BF16 22B 46 GB 53 GB Workstation
Dev and Distilled (FP8) FP8 22B 30 GB 34 GB Workstation

Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more.

The license in plain words

The custom license for LTX-2 and LTX-2.3, dated 5 January 2026. Organizations with annual revenue under US$10 million can use, modify and host the models for any purpose, subject to a list of banned uses. Organizations at or above that figure need a paid commercial agreement. LTX-2.5 uses the newer LTX-2.x license instead.

  • Free commercial use only below a revenue cap
  • Acceptable-use policy applies

More about LTX-2 Community

Frequently asked questions

What is LTX-2.3?
LTX-2.3 is the March 2026 version of LTX's open video model. It makes video with matching sound from text, an image or audio, and it can make vertical 1080×1920 video. LTX-2.5 has replaced it, but it is still widely downloaded. It is free for organizations earning under US$10 million a year.
Can I use LTX-2.3 commercially?
Yes, with conditions. LTX-2.3 is released under the LTX-2 Community, which allows commercial use but adds extra rules: Free commercial use only below a revenue cap; Acceptable-use policy applies. Read the license text before you ship a product.
Is LTX-2.3 open source?
LTX-2.3 is open-weight, not open source in the strict sense. You can download the model files, but the LTX-2 Community adds conditions that the Open Source Initiative definition does not allow.
Can I run LTX-2.3 on my own computer?
Only on a workstation. The smallest variant, Dev and Distilled (FP8), needs about 34 GB of memory at 4-bit precision, so plan for a 48 to 96 GB GPU, several graphics cards, or a Mac with plenty of memory.
When was LTX-2.3 released?
LTX-2.3 was released by LTX (Lightricks) on March 5, 2026.

Official links

Sources

Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.

Release date and facts last verified November 11, 2026.

Related models

Model Released Parameters Context License Smallest variant needs
LTX-2.5 LTX (Lightricks) 22B – LTX-2.x Community (Conditional) 16 GB Laptop or phone
Wan 2.2 Alibaba Wan 5B – 27B 14B active – Apache 2.0 (Permissive) 24 GB Gaming GPU
Cosmos 3 NVIDIA 3.86B – 64.6B – OpenMDW-1.1 (Permissive) 11 GB Laptop or phone
MiniMax H3 MiniMax 33.1B – MiniMax H3 Community (Conditional) 166 GB Big-memory desktop