# Lightricks (LTX)

> Lightricks, maker of Facetune and the open-weight LTX video models: LTX-2.5's audio and 4K, its revenue-threshold license, and how it compares.

Source: https://metavert.io/lightricks  
Published: 2026-10-21  
Updated: 2026-10-21

**Lightricks** is a Jerusalem-based creative software company, known first for the Facetune photo app and now for LTX, its family of open-weight video generation models. Its current model, LTX-2.5 (August 11, 2026), is a 22-billion-parameter model that generates video and audio together, and as of October 2026 it is the most capable open-weight video model with sound whose license applies in the United States and Europe. Lightricks pairs the models with a commercial product, LTX Studio, and a license that is free below $10 million in annual revenue.

### From Facetune to LTX

Five Hebrew University PhD students, Zeev Farbman, Nir Pochter, Yaron Inger, Amit Goldstein and Itai Tsiddon, founded Lightricks in January 2013. Facetune, released that March, became Apple's most downloaded paid app of 2017, and the company followed with Videoleap and Photoleap. It raised $135 million at a $1 billion valuation in 2019 and $130 million at $1.8 billion in 2021. Lightricks moved into generative video with LTX Studio, an AI video creation platform launched in February 2024 and opened to the public that August, and LTX Video, a 2-billion-parameter open-weight model released in November 2024. LTX-2, which generates 4K video with synchronized sound, followed in October 2025 and was open-sourced in January 2026; its technical report describes two linked transformer streams, 14 billion parameters for video and 5 billion for audio, connected by cross-attention in both directions.

### LTX-2.5

- **Audio and video together.** One model generates synchronized sound with the picture.
- **Multi-shot.** Several shots in a single pass.
- **Resolution.** Defaults to 1024×1536 at 24 frames per second, with a documented 4K UHD path (3840×2176) through spatial and temporal upscalers.
- **Pipelines.** Keyframe-guided generation, audio-to-video, IC-LoRA control and dubbing, plus a distilled version that runs in far fewer steps.
- **Hardware.** The components total tens of gigabytes; fp8 and other quantized formats with CPU or disk offloading fit smaller GPUs. ComfyUI templates were available at launch.

The weights are gated on Hugging Face: users sign in and accept the license before downloading.

### The License

LTX-2.5 ships under the LTX-2.x Community License Agreement, dated August 11, 2026. It is royalty-free and worldwide for most users, but it is not open source. Entities with annual revenue of at least $10 million need a paid license for any use beyond narrow non-commercial purposes such as evaluation. The license also bars use in products that directly compete with Lightricks' own without a separate agreement, limits using the model to train other models commercially, requires fine-tuned derivatives to stay under the same license, and forbids circumventing its watermarking and provenance features. Lightricks claims no rights in generated output. For startups and creators under the threshold it is effectively free; larger companies need a commercial deal.

### How It Compares

Among open-weight models, LTX-2.5's closest rival is Alibaba's [Wan](https://metavert.io/wan) 2.2, which is a year older, silent and limited to five-second clips at 720p, but carries the far more permissive Apache-2.0 license. Tencent's [HunyuanVideo](https://metavert.io/hunyuanvideo) 1.5 is efficient but excluded from the EU, UK and South Korea by its license. [MiniMax](https://metavert.io/minimax)'s H3, whose base weights were released in August 2026, ranks far higher on the leaderboards, but its license excludes the US as well as the EU, UK and South Korea. Against the closed leaders, LTX trails: on Artificial Analysis's text-to-video-with-audio board on October 20, 2026, the hosted LTX-2.5 Fast and Pro ranked 23rd and 24th, behind Wan 3.0, [Seedance](https://metavert.io/seedance) 2.5 and others, but they are among the few entries whose weights anyone can download. For teams in the US or Europe that want an actively developed open model with sound, it is the strongest option in late 2026. See [LTX vs Wan](https://metavert.io/compare/ltx-vs-wan), [open-weight models](https://metavert.io/open-weight-models), [text-to-video](https://metavert.io/text-to-video) and [generative video](https://metavert.io/generative-video).

## Related Topics

- [LTX vs Wan](https://metavert.io/compare/ltx-vs-wan) — Open-weight video models for commercial use
- [Wan](https://metavert.io/wan) — Alibaba's open and closed video models
- [Open-Weight Models](https://metavert.io/open-weight-models) — Models you can run yourself
- [Text-to-Video](https://metavert.io/text-to-video) — How video generation models work
- [Generative Video](https://metavert.io/generative-video) — The broader field
- [HunyuanVideo](https://metavert.io/hunyuanvideo) — Tencent's open-weight video models
- [Diffusion Transformers (DiT)](https://metavert.io/diffusion-transformers) — The architecture behind most video models
- [Generative Audio](https://metavert.io/generative-audio) — Sound generated with the picture

## Further Reading

- [Lightricks/LTX-2.5](https://huggingface.co/Lightricks/LTX-2.5) — Hugging Face; the gated model repository
- [LTX-2 repository](https://github.com/Lightricks/LTX-2) — Lightricks; LTX-2.5 weights, pipelines and quantization
- [LTX-2.x Community License Agreement](https://raw.githubusercontent.com/Lightricks/LTX-2/main/LICENSE-2_x) — Full license text, dated August 11, 2026
- [LTX-2 technical report](https://arxiv.org/abs/2601.03233) — Lightricks, 2026; joint audio-video architecture
- [Lightricks](https://en.wikipedia.org/wiki/Lightricks) — Wikipedia; company history