# Lightricks account plan

> Lightricks (Jerusalem) makes the Facetune consumer apps; its LTX open world-model unit was spun off on July 8, 2026.

## At a glance

Lightricks is a Jerusalem company founded by Hebrew University PhD students; it makes consumer mobile creative apps (Facetune, Videoleap, Photoleap) and, through its LTX unit, open-weights video and world models. LTX says it was funded by the profits of the apps, not by external capital. On July 8, 2026 LTX was spun off as an independent open world models company led by co-founder and CEO Zeev Farbman, while Lightricks keeps the consumer apps under General Manager Asaf Porat. The summaries give no revenue or headcount figures for the company.

## Business units

- **Lightricks consumer apps (Facetune)**: The mature B2C business: mobile creative tools such as Facetune, with Videoleap and Photoleap also named. It continues to be operated by Lightricks after the LTX spin-off, with Asaf Porat as General Manager. CEO Zeev Farbman called the combined group a "very profitable existing mobile business" next to a loss-making moonshot.
- **LTX (open world models)**: Open-weights video, audio and world models (LTX-2, LTX-2.3, LTX-2.5), sold through the LTX API and LTX Studio, with a free tier for organizations under $10M in annual recurring revenue. Spun off from Lightricks on July 8, 2026 and led by Zeev Farbman. It reports more than 33 million downloads of its models and targets film and VFX, real-time applications and robotics.

## Strategic priorities

- **Open world models as a standalone company**: LTX describes itself as an open world models company that builds models to understand, shape and simulate virtual and physical worlds, for filmmaking, gaming, robotics and manufacturing. The spin-off was announced on July 8, 2026.
- **Open weights with commercial licensing**: Models are released as open weights; use is free below $10M in annual recurring revenue, with a licence above that. The pitch to enterprises is that they run, fine-tune and extend models on their own hardware so IP stays in their environment.
- **Efficiency through compressed latent space**: LTX bets on an extremely compressive latent space so models need fewer tokens, cost less to train and run on consumer GPUs. It claims roughly one-eighth the cost and one-tenth the render time of comparable models.
- **Enterprise adoption**: LTX says demand comes from Fortune 500 companies, VFX houses, production studios and early-stage robotics startups. It cites up to 85% faster turnaround in pre-visualization workflows for early adopters.
- **Roadmap to local and archival video**: The February 2026 letter lists upscaling archival video to 4K and beyond, full video inference on smartphones and laptops without cloud dependency, deeper enterprise integrations and broader multimodal inputs.

## Latest results

On August 11, 2026 LTX released LTX-2.5, which it calls its most capable open-weights world model, and says it is already the most used open world model with more than 33 million downloads. Its July 8, 2026 announcement had cited 22M+ downloads on Hugging Face since the January 2026 release of LTX-2. LTX-2.5 adds native multishot generation, a custom Gemma 4 language backbone and a checkpoint tuned for physical AI and robotics. No revenue or profit figures for the period are stated in the documents.

## Investments and deals

- **LTX-2 development**: Farbman says LTX-2 was built for $150 million with a few dozen engineers and had more than 3 million downloads in its first month.
- **Spin-off of LTX**: LTX was split from Lightricks on July 8, 2026, after Facetune and LTX were first divided into two business units in a letter dated February 23, 2026.
- **Partnerships for LTX-2.5**: Launch partners are Asteria, ComfyUI, Markov Robotics and Reactor; with NVIDIA, LTX-2.5 is optimized for RTX GPUs and DGX Spark.
- **Mixture-of-experts model**: On the Cognitive Revolution podcast, Farbman said LTX planned a mixture-of-experts model of around 100 to 200 billion parameters alongside its dense models.

## Challenges and risks

- **Proving a business model for open models**: Farbman said "the jury is still out on how much place there is for open models in the West" and that investors will ask who is going to pay for them.
- **Compute constraints**: He said compute is the company's constraint because it funded model development with profits from its mobile apps, against much larger players.
- **Physics and controllability**: He named physics fidelity and creator control over every nuance as the main gap between what creators expect and what the tools deliver.
- **Long-context world consistency**: Autoregressive world models hold roughly 30 to 60 seconds of context, which he said is not enough for real games, so he did not expect games on these systems in the next quarter or two.
- **Competition from closed labs**: Farbman argues closed labs aim for a "toll road" model to justify trillion-dollar valuations and that this squeezes builders who want weights, fine-tuning and predictable costs.

## What leadership says

> "By keeping LTX open, we let teams own their hardware, their IP, and their model." — Zeev Farbman, Co-founder and CEO, LTX (Lightricks), Introducing LTX-2.5, LTX newsroom, August 11, 2026

> "Every dollar we've put into LTX has been earned from users who love our tools and trust our brand." — Zeev Farbman, Co-founder and CEO, LTX (Lightricks), LTX and the Democratization of Creative Production, LTX newsroom, February 23, 2026

> "the jury is still out on how much place there is for open models in the West" — Zeev Farbman, Co-founder and CEO, Lightricks / LTX, Interview with Superintelligence, March 2, 2026


## Where a supplier can help

- **Enterprise pre-visualization and ad production tooling**: LTX targets film, VFX, gaming and advertising teams and cites up to 85% faster pre-visualization turnaround (enterprise adoption priority). Suppliers of production pipelines, asset management or rights tooling could sell into studios fine-tuning on their own assets.
- **GPU and local-inference infrastructure**: LTX-2.5 is optimized for local inference on NVIDIA RTX GPUs and DGX Spark and is meant to run on hardware customers own (efficiency priority; compute-constraint challenge). Hardware, hosting and inference-optimization vendors fit.
- **Go-to-market for a newly independent company**: After the July 8, 2026 spin-off, LTX is a standalone company that must prove who pays for open models (business-model challenge). Sales, licensing, billing and enterprise-onboarding suppliers could help convert the free tier into multi-year licences.
- **Robotics and simulation data partners**: LTX is adding a checkpoint for physical AI and robotics and works with Markov Robotics (roadmap and investments). Domain data providers and simulation vendors could supply the fine-tuning data it says teams need.

_Tamtam's reading of the priorities and challenges above, not the company's own words._

## Key documents

### [Exclusive Interview with LTX CEO, Zeev Farbman (Superintelligence newsletter)](https://read.getsuperintel.com/p/an-exclusive-interview-with-ltx-ceo-zeev-farbman)

Executive talk · Zeev Farbman, Co-founder and CEO, Lightricks / LTX · 2026-03-02

- Lightricks (Jerusalem) began as a mobile creative-tools company: the founders finished their PhDs at Hebrew University in 2012 and set out to become "the Adobe of mobile", starting with Facetune; Videoleap and Photoleap followed.
- The move into AI models came from an economic problem: closed model providers would not hand over weights and API pricing "just didn't make sense" for the products and mobile deployments Lightricks wanted to build.
- Farbman's thesis is that the largest Western AI labs subsidise builders today to lock them in later, as Microsoft did with operating systems; open, efficient, IP-safe models are the alternative Lightricks decided to build (LTX-2 with open weights and training code).
- Lightricks is splitting its profitable consumer app business from the loss-making LTX "moonshot" so investors can judge each on its own merits; he calls it both an operational and a capital-markets decision.
- Technical bet: an extremely compressive latent space (fewer tokens, lower training cost, faster inference) lets LTX-2 run on a consumer GPU such as a 3090 card that is more than five years old.
- Developer strategy: integrations such as an MCP integration so LTX can act as a rendering engine inside tools like Blender, and a vision of diffusion models as local rendering engines for AR/VR, plus fine-tuned world models for robotics simulation.
- Candid caveat: "the jury is still out on how much place there is for open models in the West"; investors will ask who pays for open models.

### [AI:AM Highlights (The Cognitive Revolution): Zeev Farbman on open world models, 'the CapEx trap' and LTX video gen](https://www.cognitiverevolution.ai/ai-am-highlights-exploring-the-j-space-ai-superforecasters-sambanova-s-chips-ltx-video-gen/)

Podcast · Zeev Farbman, Co-founder and CEO, LTX (Lightricks) · 2026-07-09

- Farbman says video models are becoming "a backbone of what we call now like world models": like an LLM predicting the next token, a world model predicts "the next moment", including how the world looks, sounds and what actions are possible.
- LTX 2.3 was released about a quarter before the conversation; the next model was due "really soon", with a mixture-of-experts model of roughly 100 to 200 billion parameters planned beside the dense models.
- Compute is the company's constraint: it funded the model using profits from its mobile content-creation apps and competes with much bigger players.
- Real-time use cases are close: avatars are "extremely easy" with latency below a second, and robotic-arm use cases need little context; games are not, because world consistency (the coin in the drawer example) needs context compression nobody has yet.
- Business-model argument: closed labs try to build a "toll road" to justify huge valuations; LTX offers its model free below a $10 million revenue threshold, then a predictable multi-year licence above it.
- Fine-tuning range is wider than for LLMs: VFX and animation on a specific franchise, in-betweening, avatars for marketing, computational photography (low-light denoising, higher dynamic range) and fluid-dynamics simulation.
- Gaps he names: physics fidelity and controllability for creators; he sees edge compute as the most underpriced variable.

### [LTX and the Democratization of Creative Production (Zeev Farbman, LTX newsroom)](https://ltx.io/newsroom/ltx-and-the-democratization-of-creative-production)

Report · Zeev Farbman, Co-founder and CEO, Lightricks / LTX · 2026-02-23

- Lightricks divided into two business units, Facetune (the mature B2C business) and LTX (generative AI creative production), after deciding at the end of 2025 to restructure internally and accelerate investment in LTX.
- Farbman states that LTX "has been fully self-funded by the success and profitability of our existing business".
- LTX-2 was built "for $150 million with a few dozen engineers" and had more than 3 million downloads on Hugging Face in its first month; it was released as an open-weights model with commercial licensing.
- Claimed economics against closed models such as Sora: similar output quality at approximately 1/8 the cost and roughly 1/7 the rendering time, with local execution on high-end consumer GPUs.
- Enterprise pitch: the LTX API and LTX Studio, with early adopters reporting up to 85% faster turnaround in pre-visualization workflows; IP-sensitive sectors named are film, gaming and advertising.
- Roadmap: upscaling archival video to 4K and beyond, full video inference on phones and laptops, deeper enterprise integrations and broader multimodal inputs.

### [Introducing LTX as an Open World Models Company (LTX newsroom)](https://ltx.io/newsroom/introducing-ltx-as-an-open-world-models-company)

Report · 2026-07-08

- LTX announced on July 8, 2026 that it is "officially spinning LTX off from Lightricks" as an independent open world models company.
- Lightricks continues to operate and develop its consumer mobile apps, including Facetune, with Asaf Porat as General Manager of Lightricks; LTX stays led by co-founder and CEO Zeev Farbman and his co-founding team.
- LTX says it has spent 18 months developing open world models and, since releasing LTX-2 in January 2026, has become the top open world model with 22M+ downloads on Hugging Face.
- Enterprise demand is described as spanning Fortune 500 companies, VFX houses, production studios and early-stage robotics startups.
- Claimed output: native 4K, synchronized audio-video, up to 50fps, at roughly one-eighth the cost and one-tenth the render time of comparable models.
- Target use cases: filmmaking and gaming, robotics and manufacturing, autonomous-vehicle edge cases, construction and planetary rovers.

### [Introducing LTX-2.5: The Open World Model for Video, Real-Time, & Physical AI (LTX newsroom)](https://ltx.io/newsroom/introducing-ltx-2-5)

Report · 2026-08-11

- LTX released LTX-2.5 on August 11, 2026, described as its most capable open-weights world model, available on Hugging Face, natively in ComfyUI from day one and through the LTX API.
- LTX says it is already the most used open world model, with more than 33 million downloads.
- New in 2.5: a diffusion video decoder, native multishot generation, a custom Gemma 4 language backbone with a prompt enhancer, "Diffusion Fidelity Rendering" in an 8x temporally compressed latent space, a pretrained checkpoint for physical AI and robotics, and a better distilled model.
- Launch partners: Asteria (film and video), ComfyUI, Markov Robotics (robot training) and Reactor (real-time generative video); NVIDIA optimised local inference for RTX GPUs and DGX Spark.
- Licensing: open weights, free to use for organizations under $10M in annual recurring revenue; managed generation through the LTX API.
- CEO Zeev Farbman: "By keeping LTX open, we let teams own their hardware, their IP, and their model."

## Questions about Lightricks

### What does Lightricks do?

Lightricks is a Jerusalem company that makes consumer mobile creative apps such as Facetune, Videoleap and Photoleap. Its LTX unit builds open-weights video and world models.

### Is LTX still part of Lightricks?

No. LTX announced on July 8, 2026 that it was spinning off from Lightricks as an independent open world models company. Lightricks continues to run its consumer apps, including Facetune.

### Who leads Lightricks and LTX?

Zeev Farbman, co-founder and CEO, continues to lead LTX with his co-founding team. Asaf Porat is General Manager of Lightricks.

### How is LTX funded?

LTX says it has been fully self-funded by the profits of the Lightricks consumer apps, and that LTX-2 was built for $150 million with a few dozen engineers.

### What is LTX's strategy?

LTX releases open-weights world models, free below $10M in annual recurring revenue, and sells enterprise access through an API and LTX Studio, aiming at film, real-time applications and robotics.

_Written by Tamtam from the company's own published documents listed above, read in full. Figures are as those documents state them, for the periods they name. (2026-11-19)_
