Decart Review: Key Features and Pros&Cons

Name: Decart
Author: Decart

What it is:Decart is a vertically integrated AI research lab building the world's fastest real-time generative video and multimodal models.
Best for:AI developers building realtime video apps, Gaming companies exploring generative worlds, E-commerce platforms needing virtual try-on
Pricing:Starting from $0.04/sec (480p), $0.08/sec (720p)
Rating:85/100Very Good

Visit website

Reviewed byMaxim Manylov·Web3 Engineer & Serial Founder

Company Overview

Decart is an artificial intelligence research lab developing generative models at the highest speed level, which are specifically designed for real-time video modification and multi-modal artificial intelligence. The main focus of Decart is to optimize every single layer of their generative AI systems for real-time use, by developing proprietary infrastructure and efficiency-based architectures.

Active

📍Tel Aviv, Israel

📅Founded 2023

🏢Private

TARGET SEGMENTS

EnterpriseDevelopersConsumersGaming

Key Metrics

📊

$100M

Series B Funding

📊

$153M+

Total Funding

👥

1M+ in 72 hours

Oasis Launch Users

📊

$0.25 per hour vs $100 industry standard

Video Generation Cost

📊

10x more efficient than competitors like OpenAI's Sora

Model Efficiency

Credibility Rating

85/100

Excellent

Decart has demonstrated high levels of technical credibility via proprietary AI infrastructure, showing up to a 10 times improvement in efficiency; supported by tier-1 investor backing, and demonstrating solid proof of product-market fit via rapid user growth.

BREAKDOWN

Product Maturity80/100

Company Stability90/100

Security & Compliance75/100

User Reviews85/100

Transparency85/100

Support Quality80/100

TRUST SIGNALS

Backed by Sequoia Capital and Benchmark10x efficiency advantage over competitorsOasis reached 1M users in 72 hoursProfitable on early products while reinvesting in R&DMultimillion-dollar enterprise customer by month 3

Company History

2023

Company Founded

Decart was officially registered by Dr. Dean Leitersdorf and Moshe Shalev on September 7th, 2023. Both founders have extensive experience in artificial intelligence and were part of military cyber units.

2023

Seed Round & Sequoia Backing

Emerged from stealth mode with a $21 million seed round led by Sequoia Capital, demonstrating unprecedented efficiencies in real-time AI training and inference.

2023

First Enterprise Customer

By December 2023, Decart had attracted an enterprise customer worth millions of dollars via its GPU efficiency optimization product.

2024

Oasis Launch

Launched Oasis, the world’s first fully playable real-time AI generated game world, which reached 1 million users within 72 hours.

2024

Series A Funding

Benchmark led the Series A funding of $32 million to scale Decarts’ infrastructure and develop next-generation real-time AI experiences.

2024

LSD Product Release

Developed LSD (Live-Stream Diffusion), the first model capable of generating video-to-video in less than 100 milliseconds in real-time.

2025

Series B Funding

Secured a massive $100 million funding round to accelerate its mission to power the world's AI-generated pixels across multiple products and platforms.

2025

Lucy Product Launch

Launched Lucy, allowing the first-ever real-time video editing of surgical footage with frame-by-frame control and localized transformations.

Key Executives

Dr. Dean Leitersdorf— CEO & Co-founder: Served 13 years in the Israeli Unit 8200 (cyber intelligence) in various roles. Has academic expertise in AI and cyber, and prior to founding Decart he completed his postdoctoral work in Singapore. Co-founded the non-profit organization StartAch, which develops software for Israeli non-profits.
Moshe Shalev— Co-founder: Background in the Israeli military with knowledge in cyber operations. Began collaborating with Leitersdorf immediately following Operation Guardian of the Walls in 2021, realizing the synergy of combining Leitersdorfs academic and product based thinking.

Key Features

✨

Real-Time Video Generation

Video-to-video conversion under 100 milliseconds using Live-Stream Diffusion technology to enable live modification of video streams.

✨

Open-World Game Generation

Instead of pre-building environments, Oasis creates a new game world in real-time, allowing for limitless procedural generation of gameplay.

✨

Extreme Efficiency

A proprietary architecture that is 10 times more efficient than its competitors, costing $0.25 per hour for generating videos versus the $100 cost of all other options.

✨

Live-Stream Diffusion

MirageLSD allows for limitless real-time video generation with zero latency for use in livestream applications and live interactions.

✨

Real-Time Surgical Video Editing

The Lucy system provides frame-by-frame control and local video manipulation capabilities for surgical and commercial video usage.

✨

Multi-Cloud AI Infrastructure

A proprietary training framework designed to optimize performance in multi-cloud configurations by optimizing data flow and GPU cluster configuration across multiple cloud providers.

✨

Low-Latency Inference

End-to-end optimizations from systems to model architectures allow for extremely fast inference rates for real-time interactive applications.

Tech Stack

Infrastructure

Multi-cloud training infrastructure with GPU clusters distributed globally; optimized for high-performance data storage and transfer using Backblaze for cost-effective video data management.

Technologies

PythonPyTorchCUDAKubernetes

Integrations

Multi-cloud deploymentGPU cluster orchestrationAPI access for developers

AI/ML Capabilities

Proprietary generative AI models including Mirage (livestream diffusion), Oasis (real-time world generation), LSD (live-stream diffusion with sub-100ms latency), and Lucy (real-time surgical video editing) built on deep expertise in optimization and low-level systems.

Based on company website, case studies, and official product announcements

Use Cases

Content Creators & Streamers

Real-time video manipulation for livestreams that can instantly change scenes, apply filters, and modify worlds without requiring expensive rendering equipment.

Game Developers

Create endless open world games utilizing procedurally generated environments which will reduce development time and required storage space relative to pre-designed levels.

Enterprise Video Applications

Real-time video editing for corporate communication, product placement, and video enhancement at a fraction of the price of the same processes via traditional methods (i.e., 400 times less).

Surgical & Medical Video Teams

Frame-by-frame video editing of surgical video utilizing real-time control capabilities to enhance training, documentation, and visualization without having to process the video offline.

AI Infrastructure Teams

Optimizing multi-cloud GPU to solve training efficiency bottlenecks with proprietary architecture that achieves 10x greater efficiency than industry standards.

AR/VR Application Developers

Real-time video manipulation for creation of immersive experiences with immediate environment morphing and low-latency visual updates

NOT FORLatency-Sensitive Financial Trading

Not recommended - While efficient, optimized primarily for visual AI generation rather than sub-millisecond trading operations

NOT FOROffline Batch Processing Only

Not recommended - Core value proposition is real-time performance and batch processing does not leverage primary technical advantages

NOT FORBudget-Constrained Individuals

Not recommended - Enterprise-grade infrastructure and advanced capabilities position pricing for commercial/professional users, rather than individual consumers

Pricing

Pricing information with service tiers, costs, and details
☐Service	$Cost	ℹDetails	🔗Source
Lucy Pro T2V	$0.04/sec (480p), $0.08/sec (720p)	Text-to-video quality	Official pricing page
Lucy Pro I2V	$0.04/sec (480p), $0.08/sec (720p)	Image-to-video quality	Official pricing page
Lucy Pro V2V	$0.15/sec (720p)	Highest quality video editing	Official pricing page
Lucy Fast V2V	$0.04/sec (720p)	Faster, lower-cost video editing	Official pricing page
Lucy Restyle V2V	$0.01/sec (720p)	Style-driven video restyling	Official pricing page
Lucy Motion	$0.03/sec (720p)	Trajectory-based motion control	Official pricing page
Lucy Pro T2I	$0.02 (720p)	Text-to-image generation	Official pricing page
Lucy 2 Realtime	$0.02/sec (720p)	Character transform + live editing	Official pricing page
Avatar Live	$0.01/sec (720p)	Audio-driven talking avatars	Official pricing page

Lucy Pro T2V$0.04/sec (480p), $0.08/sec (720p)

Text-to-video quality

Official pricing page

Lucy Pro I2V$0.04/sec (480p), $0.08/sec (720p)

Image-to-video quality

Official pricing page

Lucy Pro V2V$0.15/sec (720p)

Highest quality video editing

Official pricing page

Lucy Fast V2V$0.04/sec (720p)

Faster, lower-cost video editing

Official pricing page

Lucy Restyle V2V$0.01/sec (720p)

Style-driven video restyling

Official pricing page

Lucy Motion$0.03/sec (720p)

Trajectory-based motion control

Official pricing page

Lucy Pro T2I$0.02 (720p)

Text-to-image generation

Official pricing page

Lucy 2 Realtime$0.02/sec (720p)

Character transform + live editing

Official pricing page

Avatar Live$0.01/sec (720p)

Audio-driven talking avatars

Official pricing page

💡Pricing Example: 30-second realtime Lucy 2 session

Lucy 2 Realtime$0.60

30 x $0.02/sec

5-second Lucy Pro V2V batch (720p)$0.75

5 x $0.15/sec

Competitive Comparison

Feature	Decart	Runway ML	Pika Labs	Kling AI
Core Functionality	Real-time video editing	Text-to-video	Text-to-video	Text-to-video
Realtime Generation	Yes	No	No	No
Video Editing Models	Yes (V2V/I2V)	Limited	No	Limited
Image Models	Yes	Yes	Yes	Yes
API Access	Yes	Yes	Yes	Yes
Pay-as-you-go Pricing	Yes ($0.01-$0.15/sec)	Yes	Yes	Yes
Free Tier	No	Limited	Yes	No
Enterprise Features	Early stage	Yes	No	No
Realtime Cost Efficiency	10x improvement
Multimodal Support	Yes (audio avatars)	Limited	No	No

Core Functionality

DecartReal-time video editing

Runway MLText-to-video

Pika LabsText-to-video

Kling AIText-to-video

Realtime Generation

DecartYes

Runway MLNo

Pika LabsNo

Kling AINo

Video Editing Models

DecartYes (V2V/I2V)

Runway MLLimited

Pika LabsNo

Kling AILimited

Image Models

DecartYes

Runway MLYes

Pika LabsYes

Kling AIYes

API Access

DecartYes

Runway MLYes

Pika LabsYes

Kling AIYes

Pay-as-you-go Pricing

DecartYes ($0.01-$0.15/sec)

Runway MLYes

Pika LabsYes

Kling AIYes

Free Tier

DecartNo

Runway MLLimited

Pika LabsYes

Kling AINo

Enterprise Features

DecartEarly stage

Runway MLYes

Pika LabsNo

Kling AINo

Realtime Cost Efficiency

Decart10x improvement

Runway ML—

Pika Labs—

Kling AI—

Multimodal Support

DecartYes (audio avatars)

Runway MLLimited

Pika LabsNo

Kling AINo

Competitive Position

vs Runway ML

Decart excels in real-time video editing and efficiency (10X cost reduction), while Runway focuses on high-quality text-to-video generation. Decart's infrastructure enables lower-cost realtime applications

Decart leads in real-time/low-latency use cases; Runway better for cinematic offline video generation

vs Pika Labs

Pika emphasizes consumer-facing text-to-video with free tier access, while Decart targets enterprise developers with API-first realtime models. Decart's pricing is usage-based with no subscriptions

Pika for quick prototypes; Decart for production real-time applications

vs Kling AI

Both offer advanced video models but Decart differentiates with real-time capabilities and proprietary efficiency infrastructure. Kling focuses more on longer-form video generation

Decart for interactive/live experiences; Kling for narrative video content

Pros Cons

Pros

10X training efficiency — substantially lower AI model costs than industry standards
Real-time video generation — first playable AI-only video game demonstrates capability
Pay-as-you-go pricing — no subscriptions or minimum spend requirements
Early profitability — achieved millions in revenue within months of launch
Sequoia Capital backing — strong validation from top-tier VC
Proprietary infrastructure — enables realtime models at startup scale
Broad model coverage — realtime, batch video, and image generation APIs

Cons

Early stage company — launched from stealth in Oct 2024, limited production track record
High real time cost – Between $.02 and $.15/sec – may be too costly to run a session for an extended period of time $.03 per hour
No Free Tier – The company is a pay as you go type which means that customers have to commit to spend money with them before they begin using their products
Limited enterprise features – A young company does not have the mature single sign-on or audit log features that are common in the enterprise space
Narrow Focus – The company has focused their efforts into developing video and real time solutions as opposed to developing general purpose artificial intelligence platforms
Requires Cloud GPU – In order to deploy Lucy 2 real-time models, you need to have access to Nvidia or Amazon/Google cloud based infrastructure
Documentation Visibility Limited – Pricing information is readily available; however, the complete documentation for Lucy 2 is not visible from the pricing page alone

Best For

AI developers building realtime video apps — Real Time Models Optimized For Live Editing And Interaction – Lucy 2 real time models were designed with the ability to allow users to interact with the data in a real time manner
Gaming companies exploring generative worlds — Demonstrates Ability To Build Open World Generative Models – Lucy 2 is the first generative model that was able to build open world models in a real time manner
E-commerce platforms needing virtual try-on — Character Transformation At Scale – With Lucy 2's ability to perform transformations on characters in a real time manner, companies now have the ability to transform their customer service experience at a large scale
Video production teams requiring fast editing — Multiple V2V Model Types Available From $.01-.15 Sec – Depending upon the intended use case, multiple types of video-to-video models are available at varying price points
AI research labs optimizing training costs — Platform Reduces Training Expenses By 10X – Lucy 2's platform enables companies to train models 10 times faster than other platforms, resulting in a 10 fold reduction in the expense of training those models

Not Suitable For

Budget-constrained developers — No Free Tier Available; Real-Time Costs Add Up Quickly – As a result of the real time nature of the product, costs add up very quickly and there is currently no free tier offered by the company
Static image-only generation needs — Video Focused Platform – The platform is video-centric so if you are looking for a solution that focuses on text-to-image, consider using a platform that specializes in text-to-image such as Midjourney
Non-technical content creators — API-First Developer Platform Without No-Code UI – While the platform is designed for developers and offers an API, it also does not offer a no code user interface. Therefore, developers will need to create applications that utilize the API or try a platform that offers a web-based interface such as Runway ML
Enterprises requiring mature compliance — Early Stage Company Does Not Have SOC2 Or Single Sign-On – As a early stage company, the platform does not yet support SOC2 or single sign-on SSO. Therefore, if this is something that is required for your organization, you may want to consider using a more established provider such as AWS Bedrock

Limits Restrictions

Pricing Model: Pay-as-you-go only, no subscriptions or free tier
Realtime Costs: $0.01-$0.02/sec (720p), ~$3/hour sustained generation
Batch Video Costs: $0.01-$0.15 per generated second
Image Generation: $0.01-$0.02 per 720p image
Minimum Spend: No minimum but usage-based billing only
Infrastructure: Optimized for Nvidia GPUs + AWS/Google Cloud
Model Availability: Video-focused (T2V/I2V/V2V) + realtime + images
Geographic Availability: Cloud-based, likely global but US/EU focused

Security & Compliance

Early Stage InfrastructureProprietary platform achieves 10x efficiency with cluster training reliability.

Cloud InfrastructureOptimized for AWS, Google Cloud, and Nvidia GPUs with standard cloud security.

API SecurityStandard API authentication required for pay-as-you-go access.

Enterprise ReadinessEarly profitability with paying enterprise customers; security features developing.

Data ProcessingVideo/image generation processes customer inputs; standard retention policies apply.

Customer Support

Channels

Pricing and getting started guides availabledecart.ai contact formDirect support for paying customers

Hours: Business hours (assumed)
Response Time: Not publicly specified; enterprise customers receive priority
Satisfaction: N/A - early stage company
Specialized: Dedicated support for enterprise customers
Business Tier: Priority access for paying enterprise accounts

Support Limitations

•Limited public support channels for non-customers

•No 24/7 support mentioned

•Documentation still developing