Draft:Decart AI
From Wikipedia, the free encyclopedia
Decart AI Ltd. is an artificial intelligence startup and research lab developing real-time models for image, video, audio, and multimodal applications. The company is engaged in expanding the boundaries of interactive artificial intelligence, with an emphasis on transforming live content into dynamic media.[1]
Submission declined on 10 November 2025 by Theroadislong (talk).
Where to get help
How to improve a draft
You can also browse Wikipedia:Featured articles and Wikipedia:Good articles to find examples of Wikipedia's best writing on topics similar to your proposed article. Improving your odds of a speedy review To improve your odds of a faster review, tag your draft with relevant WikiProject tags using the button below. This will let reviewers know a new draft has been submitted in their area of interest. For instance, if you wrote about a female astronomer, you would want to add the Biography, Astronomy, and Women scientists tags. Editor resources
|  |
Comment: To the reviewer: This article comply with the specific notability guideline for organizations and companies (WP:COMPANY): 1) WP:ORGIN: This company has attracted the notice of reliable sources unrelated to the organization or product
2) WP:ORGCRIT This company is notable because it has been the subject of significant (WP:ORGDEPTH) coverage in multiple reliable secondary sources that are independent of the subject (WP:ORGIND).
3) The company has a valuation of $3.1 billion, making Decart a unicorn.
There is no PROMOTION in this article.
Thenks - Ovedc (talk) 06:51, 10 November 2025 (UTC)
 | |
| Founded | 2023 |
|---|---|
| Founders | Dean Leitersdorf (CEO) and Moshe Shalev (VP of Product) |
| Headquarters | Tel-Aviv |
Number of employees | 60 (2025) |
| Website | decart |
Decart is known for developments such as Oasis – an interactive open world model generated in real time by AI, and Live-Stream Diffusion (LSD) – a family of real-time video creation and editing models that enable continuous, interactive, and live restyling and editing.
History
The company was founded in September 2023 by Dean Leitersdorf (CEO) and Moshe Shalev (VP of Product), both alumni of Unit 8200.[2]
By late 2023, Decart developed GPU optimization technology, and signed initial contracts with cloud computing providers, achieving early profitability.
In October 2024, Decart emerged from stealth and launched Oasis in collaboration with Etched. Oasis is an interactive world model, generated in real time according to user input. Oasis gained over a million users in just three days and received widespread international exposure.[3]
Following the launch, Elon Musk tweeted his admiration for the technological achievement, which helped increase international coverage of Decart.[4]
In July 2025, Decart announced a collaboration with the Technion to establish a joint research center and the Decart Excellence Program, with the aim of advancing the field of AI research in Israel.[5]
During 2025, Decart opened additional branches in San Francisco and New York. As of summer 2025, the company employs over 60 people.[6] [7]
Capital raising
Since its founding, Decart has raised $153 million in three major rounds within 11 months:
- October 2024 – $21 million seed round led by Sequoia Capital with the participation of others, including Zeev Ventures (which took part in the company's pre-seed stage).[8] [9]
- December 2024 – $32 million Series A led by Benchmark Capital, at a valuation of approximately half a billion dollars.[10][11][12]
- August 2025 – Another $100 million round led by returning investors Benchmark Capital, Sequoia Capital[13] and Zeev Ventures, as well as Valley Capital Partners, Regah Ventures and Aliya Capital Partners. The funding was done at a valuation of $3.1 billion, making Decart a unicorn.[14][15]
Products
- GPU Optimization Stack – an internal infrastructure for optimizing GPU usage, which served as the basis for initial contracts with cloud providers.
- Oasis – an interactive world model, generated in real time based on user inputs. Launched in October 2024, it was the first large-scale implementation of a live generative video model.[16]
- LSD (Live-Stream Diffusion) – an autoregressive video-to-video model that generates infinite, real-time video. launched in July 2025, it allows restyling of live video while maintaining stable motion and frame rate.[17]
- Lucy Edit – an open-weight video editing model for text-guided video editing. Launched in September 2025, Lucy Edit supports edit operations such as scene replacements, clothing and accessory changes, character changes and object insertions, while preserving composition and motion.[18]
Technology
Decart developed the Live-Stream Diffusion (LSD) architecture – a model that generates continuous, real-time video.[19]
Main principles:[20]
- Autoregressive – each frame is built on the previous frame and the text command, ensuring a stable flow and visual continuity.
- Handling error accumulation – the company introduced two mechanisms:
- Diffusion Forcing – local error correction during training, to reduce long-term drift.
- History Augmentation – Using frame history to stabilize the image and "erase" repeated errors.
- Performance – The model reaches 24 frames per second at a resolution of 432×768 with a latency of less than 40 milliseconds. The company has stated its intention to expand support to Full HD and 4K.
- Hardware efficiency – The system is written in CUDA and optimized for NVIDIA GPUs, reducing processing costs from $10–$1000 per hour to less than $0.25 per hour.
This technology enables real-time applications in the areas of gaming, streaming, advertising, and video calling, where real-time interactivity is essential.[21]
Wikipedia articles must be written neutrally in a formal, impersonal, and dispassionate way. They should not read like a blog post, advertisement, or fan page. Rewrite the draft to remove:
Instead, only summarize in your own words a range of independent, reliable, published sources that discuss the subject.
If you have a conflict of interest (e.g. you are the subject, an employee, or a relative) or are being paid to edit, you must disclose this to comply with Wikipedia's Terms of Use.