Genie 3
DeepMind's advanced world model generates interactive, physically logical 3D virtual environments in real time from textual cues, and is widely used in gaming, education, and AGI research.
What is ABot-Earth Studio?
ABot-Earth Studio is a commercial platform powered by the ABot-Earth 0.5 model, touted as the world’s first 3D-native city model that is fully trained on 3D data and ready for engineering use.world model.
In a nutshell: Upload a satellite image or a text description, and within 10 minutes, generate a kilometer-scale 3D city model using a consumer-grade graphics card. The output is 3DGS-compatible assets that can be directly imported into game engines.
It has built 3D maps covering more than 190 countries and regions, making it currently the most extensive 3D map in the world.
Key Features of ABot-Earth Studio
- Generating 3D Cities from Text and Images:Enter a satellite image or text description, and the AI will automatically generate a 3D city scene;
- Multi-planet travel:Supports realistic 3D exploration of the Earth, the Moon, Mars, and numerous cities both domestically and abroad;
- Plot Square Design:Supports custom plot editing, allowing users to generate, explore, and customize plots;
- 3DGS format output:Export native, render-ready 3DGS assets for direct import into Unity or Unreal Engine;
- Levels of Detail (LOD):With built-in depth of field for both near and far distances, it supports smooth navigation across different viewing distances without the need for post-processing;
- Automated Pipeline Delivery:End-to-end automation, from input to editable 3D assets.
ABot-Earth Studio's Core Technology
| Technical Modules | Issues Addressed | Specific Plan |
|---|---|---|
| ① 3DGS Compression-Generation Framework(Training Side) | 3DGS data contains millions of primitives and is unstructured, making it impossible to train large models directly | We pioneered a compression-to-generation framework that directly processes 3DGS point clouds, encoding high-quality real-world 3DGS scenes into a compact latent space to address the issue of disorder, enabling the model to directly ”read” large-scale 3D data. |
| ② Sliding-Window Inference Mechanism(Deduction Side) | Scenes spanning kilometers cannot be generated in a single pass | Block-based generation combined with intelligent merging of overlapping areas enables continuous construction across wide areas spanning kilometers, ensuring spatial continuity. |
| ③ Cross-domain Adaptive Module(Generation side) | There is a resolution gap between satellite imagery (2D) and 3D training data | Bridge the gap between 2D and 3D domains to ensure that text and satellite image inputs accurately drive 3D generation |
| ④ Multi-level detail decoder (LOD)(Rendering side) | The scene needs to switch seamlessly from a bird's-eye view to street-level detail | Built-in LOD ensures that the generated results feature natural depth of field, allowing for smooth navigation at any viewing distance. |
Key Differences: Instead of taking the traditional approach of ”distilling 3D structures from 2D images,”Train the model directly using 3D data...establish a native understanding of three-dimensional space and generate 3DGS-format scenes end-to-end.
Key Performance Indicators for ABot-Earth Studio
| norm | digital |
|---|---|
| Generation speed | 10 minutesGenerating 3D City Models at the Kilometer Scale (Consumer-Grade Single-Card GPU) |
| Efficiency gains | An improvement of approximately 1,000 times |
| (manufacturing, production etc) costs | Limited to traditional 3D drafting 1% |
| Coverage | Over 190 countries and regions |
| output format | 3DGS (native renderable, ready for Unity/UE) |
Use Cases for ABot-Earth Studio
- Embodied Intelligence / Robotics:Build simulation training environments in minutes to support the transfer of robot algorithms to real-world scenarios. Amap’s city-scale simulation training environment is powered by ABot-Earth 0.5 and supports urban simulation training for ”Amap Tutu,” the world’s first fully autonomous robot designed for open environments;
- Low-altitude economy:Rapidly generate 3D terrain models using satellite imagery, covering areas traditionally overlooked by surveying and mapping, and providing a map reference for areas previously without one;
- Film, Television, and Video Games:AI-generated output replaces manual work by modelers, allowing creators to focus their energy on the creative process itself;
- Emergency Response:Reconstructs a 3D environment of the disaster site within minutes, providing a scientific basis for command and dispatch;
- General 3D Content Production:Any industry that requires 3D city base maps is shifting from a ”capital-intensive” model to a ”tool-based” one.
Comparison of similar products
| comparison dimension | ABot-Earth Studio (AutoNavi) | Google Earth Studio (Google) |
|---|---|---|
| core positioning | Native 3D city world models,AI generation3D scene | Images of Earthanimationartifact |
| Technical Approach | Native 3D data training → Generate 3DGS | Based on existing Google Earth imagery → Keyframe animation |
| input method | Satellite image / Text description (generated by AI) | Manually set the camera path and keyframes |
| output format | 3DGS Editable Assets(Jump straight into Unity/UE) | Video/Animation (Export to After Effects) |
| generating capacity | AI-Generated 3D City at the Kilometer Scale in 10 Minutes | I cannot create new scenes; I can only animate existing footage. |
| Efficiency | Delivery in minutes, with efficiency increased 1,000-fold | Depends on the time required for manual production |
| (manufacturing, production etc) costs | Traditional 1% | Free (for news use); commercial use requires a license |
| 3D Availability | ✅ Native 3D, interactive and editable | ❌ This is essentially a 2.5D animated video; the 3D structure cannot be edited. |
| Coverage | Over 190 countries and regions | Global Google Earth Coverage |
A brief summary of the comparison:
Google Earth Studio is about ”making movies of the existing Earth,” while ABot-Earth Studio is about ”using AI to build an interactive Earth from scratch.” The two are not competing on the same level—the former is an animation tool, while the latter is a 3D world generation engine.
✅ The closed beta is now officially open. You can go to abot-earth.amap.com Submit an application to try it out.
If your requirement isGenerate editable 3D city assets for developmentABot-Earth Studio is currently the only native 3D solution; if you only needEarth imagery animations/video footage... Google Earth Studio remains the go-to choice.
data statistics
Related Navigation
Qwen-AgentWorld
AliQianwen, the first native language world model released on June 24, 2026, uses plain text to uniformly simulate seven major digital environments, allowing agents to practice in a "virtual world."
MIDI (loanword)
AI 3D scene generation tool that can efficiently generate complete 3D environments containing multiple objects from a single image, widely used in VR/AR, game development, film and television production and other fields.
ABot-World Studio
Gaode’s General-Purpose World Model Workshop enables users to quickly generate highly realistic AI digital worlds—capable of real-time interaction and stable operation for hours—using text and images. It can be deployed locally with a single RTX 5090, and its core model has been fully open-sourced.
Firefly Image 5
The latest generation of generative image models from Adobe supports native high-resolution, layered editing with natural language commands, allowing creators to generate, edit, and customize high-quality images that match their brand's style in a single sentence.
Dzine
AI image generation and editing platform that allows users to quickly create high-quality visual content through simple operations.
Supawork AI
Free online multifunctional AI creation platform, integrating avatar generation, resume optimization, video effects, academic assistance and other functions, without registration to use.
SellerPic
Designed for e-commerce sellers, the AI image tool can efficiently generate professional fashion model images and high-quality product images, helping sellers save costs, improve visual appeal and optimize marketing strategies.
No comments...
简体中文 
English