Home
Gaming
Decart

Decart

Decart is a research-led platform providing the first real-time generative world models that render interactive environments frame-by-frame without a traditional game engine.

Excellent for AI researchers and experimental developers exploring engine-less simulation, but currently too low-fidelity for commercial game production.

Ai Video Generation Real-Time Game Development Generative Ai Worldbuilding Simulation Ai Models

Analysis based on product data, pricing structure, traffic signals, and public user sentiment.

Visit Website

Share this tool

Facebook X

LinkedIn Reddit

WhatsApp Telegram

Who Should Use Decart?

Typical users

AI researchers, creative technologists, and forward-looking game developers interested in generative simulation.

Maturity fit

beginner

Choose this if…

You want to test the limits of real-time AI video generation
Your priority is exploring interactive world models over high-fidelity graphics
You are researching alternatives to traditional rasterization and ray-tracing engines

Skip this if…

You need high-resolution, crisp visual assets for a commercial project
Your game requires strict object permanence and long-term state consistency
You need a production-ready SDK with extensive documentation and support

About Decart

Decart is an AI lab focused on building generative world models that simulate complex environments in real-time. Their flagship model, Oasis, allows users to interact with a Minecraft-like world where every frame is generated on the fly by a neural network rather than a game engine.

Official profiles

X (Twitter) LinkedIn Discord

What it actually does

It replaces the traditional game loop—physics, rendering, and logic—with a single transformer-based model. Users provide keyboard and mouse inputs, and the model predicts and generates the resulting video frames at 20 frames per second.

What makes it different

Unlike standard AI video generators (like Sora or Runway) which are non-interactive, Decart's models are optimized for ultra-low latency. It bypasses game engines entirely, meaning the 'physics' of the world are learned from video data rather than hard-coded by developers.

Real-time interactive frame generation Zero-engine environment rendering User-input responsive simulation Open-source model weights (Oasis 500M) Minecraft-style world generation Dynamic environment morphing

Key Features

Oasis Model

A world model trained on gameplay video to simulate physics and visuals simultaneously.

Real-time Inference

Optimized architecture that allows for playable frame rates (approx. 20fps) on modern GPUs.

Interactive Latency

Designed to respond to user controls with minimal delay, mimicking a local game experience.

Engine-less Rendering

Eliminates the need for assets, textures, or lighting maps by generating them procedurally via AI.

Open Weights

Provides a smaller version of the model for the research community to inspect and run locally.

Dynamic World Building

The environment can change drastically based on user movement, as the model 'dreams' the next frame.

Pricing

Popular

Oasis Demo

Free

Free browser-based play
Access to the Minecraft-style world model
Standard latency
Limited session length

Developer/Enterprise

Custom

API access for custom world models
Higher resolution output
Dedicated compute resources
Technical support

Pricing checked 5 months ago

Pricing guidance

Best plan for most users: The free Oasis demo is the only accessible option for most users to test the technology's current state.

Free plan enough? Yes, for anyone just wanting to see how a generative world model feels to play.

Upgrade when:

When you need to integrate world models into a custom application
When you require higher resolution than the public demo
When you need to train a model on your own proprietary video data

Watch out for:

Public demo sessions are often queued or time-limited
Resolution is capped to maintain real-time performance
No save-game functionality exists in the current model

Currently positioned as a research-first lab with a free public showcase to drive interest and talent.

Pros & Cons

Strengths

Proof of concept for engine-less gaming
It demonstrates that a neural network can handle physics and rendering in one pass, which could eventually reduce game development overhead.
High interactivity
Unlike other generative video tools, the feedback loop is fast enough to feel like a game, providing a glimpse into future 'infinite' content generation.
Open research contribution
By releasing model weights and technical details, Decart allows developers to understand the underlying transformer architecture used for world modeling.

Weaknesses

Low visual fidelity
The output is often blurry, low-resolution, and contains significant visual artifacts that would not be acceptable in a modern commercial game.

Affects: Game developers and players used to HD/4K standards
Lack of object permanence
Because the model is 'dreaming' frames, objects can disappear or change shape when the user looks away and looks back, breaking immersion.

Affects: Gameplay mechanics requiring inventory or persistent structures
High compute requirements
Running these models in real-time requires massive GPU power (H100s), making it expensive and difficult to scale for mass audiences currently.

Affects: Indie developers and budget-conscious studios

Real User Sentiment

Users are generally impressed by the technical feat of real-time generation but recognize it is currently more of a 'hallucinating' toy than a stable game.

Users tend to like

The 'magic' feeling of a world being created as you move
The technical achievement of 20fps generative video
The open-source availability of the 500M model

Users commonly complain about

Visual blurriness and 'dream-like' distortions
Items disappearing from the world (hallucinations)
Limited gameplay depth beyond basic movement and interaction

Recurring tradeoffs

You trade visual clarity and consistency for the ability to have a world that is entirely generated on the fly.

Happiest users

AI enthusiasts and researchers who value the novelty of the tech over the quality of the game.

Often frustrated

Gamers expecting a functional Minecraft clone or developers looking for a ready-to-use game engine replacement.

Use Cases

AI Research

Testing how transformers handle spatial reasoning and physics.

Infinite World Exploration

Creating environments that never end and never repeat.

Rapid Prototyping

Visualizing environment concepts without building 3D assets.

Interactive Art Installations

Using user input to drive a constantly evolving visual world.

Education

Demonstrating the capabilities of generative AI to students and developers.

Frequently Asked Questions

Is Decart Oasis free to play?

Yes, Decart currently offers a free browser-based demo of Oasis. You can jump into a Minecraft-like world and interact with it in real-time without paying, though there may be a queue during high-traffic periods.

How does Decart compare to Google Genie?

Both are world models, but Decart's Oasis is currently more accessible as a playable public demo. While Google Genie was a research paper showing 11fps generation, Oasis achieves roughly 20fps, making it feel more like a functional game.

Can I build my own game using Decart right now?

Not easily. While they have released open-source weights for a smaller model, there is no public SDK or 'Unity-style' editor. It is currently a research platform rather than a commercial development tool.

What are the main limitations of the Oasis model?

The biggest issues are 'hallucinations' and low resolution. The model often forgets where objects were placed if you turn around, and the visuals are significantly blurrier than traditional games.

Does Decart integrate with Unreal Engine or Unity?

No. Decart is designed to replace the need for those engines entirely. It does not use traditional assets like meshes or textures, so it is not compatible with standard game engine workflows.

What hardware do I need to run Decart models?

To run the full-scale Oasis model in real-time, you need high-end enterprise GPUs like the NVIDIA H100. However, the smaller 500M parameter model can be run on consumer-grade hardware for research purposes.

Why trust this page?

This evaluation combines product positioning, pricing analysis, traffic and market signals, and public user sentiment into a single decision-support page. Content is generated editorially — not copied from the vendor's website.

Funding & Company

Founded

2023

Stage

Series c

Total Raised

$453M

Latest Round

Series C (May 2026)

Notable Investors

Sequoia Capital Benchmark Radical Ventures NVIDIA Zeev Ventures Adobe Ventures Toyota Ventures

Decart has raised a total of $453 million across four rounds in just over 18 months, indicating exceptionally strong investor confidence. Its latest $300 million funding round, which included strategic investment from NVIDIA, provides substantial capital to scale its real-time AI infrastructure and world-model platform, ensuring a long operational runway.

Full funding report high confidence

Market Signals & Traffic

Estimated visits, global rank, geography, traffic sources, monthly visit trends, and organic search keywords (Similarweb)—on a dedicated page built for depth and search.

Estimated visits: 320,860
Global rank: #121,752
Snapshot: Apr 2026
Traffic trend: Falling