Turn any image into an editable 3D world with Rodin WorldGen

Imagine taking a single 2D image and turning it into a fully editable 3D world you can walk through, rearrange, and export into Blender, Unity, or Unreal Engine. That’s exactly what Rodin’s new WorldGen feature does, combining AI-generated 3D objects with 3D Gaussian Splatting (3DGS) backgrounds to create rich, usable scenes from just one picture.

What is Rodin WorldGen?

Rodin WorldGen is an AI-powered tool that converts a single image into a complete 3D environment. It automatically detects objects in the image, turns them into editable 3D meshes, and builds a 3DGS background around them. The result is a scene you can explore, tweak, and export to major 3D and game engines.

Instead of just getting a single 3D model from a photo, you get an entire world: foreground objects, background geometry, lighting, and camera-ready composition, all in one workflow.

How object extraction from a single image works

WorldGen gives you two main ways to turn parts of an image into 3D objects: automatic extraction and manual selection.

Automatic object extraction

The automatic mode analyzes the image and starts extracting objects one by one. You can watch as it detects an object, generates its 3D geometry, and then applies textures. This happens iteratively until the main elements in the scene are converted into individual 3D assets.

Each extracted object becomes an editable mesh in the scene, so you can move it, resize it, or regenerate it later if you’re not happy with the result.

Manual selection with pins

If you want more control, you can add objects manually. WorldGen lets you click “Add New Object” and then use pins to mark the area of the image that should become a 3D object. You can also place pins to exclude areas that shouldn’t be part of that object.

This is especially useful when the automatic extraction groups too many things together—for example, when you want a single barrel instead of a whole cluster. You can refine the selection, isolate just what you need, and generate a clean, separate 3D asset.

Editing and refining extracted objects

Once objects are extracted, you’re not locked into the first result. WorldGen gives you several ways to refine them:

• Regenerate objects: If you don’t like how a model turned out, you can regenerate it to get a better version.

• Refine boundaries: Adjust the edges of the extracted area so the mesh more accurately matches the object in the image.

• Edit extracted image patches: You can open the “extracted image” for an object and tweak what’s included before regenerating the 3D model.

This object-centric workflow makes WorldGen useful even if you only care about building an asset library. You can batch-extract multiple props from a single image and export them as separate 3D models, similar to what other image-to-3D tools do, but with more control over each piece. If you’re interested in this broader space, it’s worth comparing with tools covered in this guide to turning images into production-ready 3D models.

SimReady: adding physical properties for robotics

WorldGen also offers a mode called SimReady. When enabled, it automatically assigns physical properties—like weights and other simulation-related attributes—to the objects it generates.

These physics-aware assets are exported in USDZ format, which is designed to support simulation data. This is especially relevant for robotics and simulation workflows, where you want to train robots or agents in realistic virtual environments without manually setting up physics for every object.

If you’re not working in robotics or physics simulations, you can simply turn SimReady off and stick to standard 3D exports.

Arranging the scene and fixing object placement

After extraction, all your 3D objects are placed in a scene where you can move, rotate, and rearrange them however you like. To help match the original photo’s layout, WorldGen includes a feature called “Fiz Correction” (a kind of physical/position correction).

This option analyzes the original image again and tries to recompose the objects in 3D so they roughly match their positions in the picture. It isn’t perfectly accurate, but it usually gets close enough to serve as a solid starting point—especially for objects that should sit on the ground or align with a floor plane.

Generating 3DGS backgrounds for full environments

Once you’re happy with your objects, you can generate the full environment using 3D Gaussian Splatting. WorldGen offers two modes:

• High-quality mode: Produces better-looking backgrounds but can take 10–20 minutes to generate, depending on the scene.

• FastGen mode: Much faster, with lower quality—useful for quick previews or blocking shots.

The result is a 3DGS background that blends with your extracted objects, creating a complete 3D world. Up close and from tricky angles, you may notice some limitations typical of splatting, but for many use cases—like establishing shots, previs, and blocking—it looks surprisingly convincing.

WorldGen’s environments can range from open outdoor locations to tighter, more enclosed spaces like streets or courtyards, and they’re detailed enough to use as base scenes for games, cinematics, or concept exploration.

AI Render: turning scenes into images and videos

WorldGen also includes an AI Render feature that lets you use your 3D scene as input for image and video generation. The idea is simple:

• You arrange your 3D objects and environment.

• The scene is passed into an image-editing model.

• You generate either still images or short videos based on that setup.

Unlike tools that only generate a flat video from text prompts, this approach gives you a real 3D scene you can keep editing. You can change object positions, adjust the composition, and then re-render new shots or motion clips, which is especially useful for planning cinematic sequences or experimenting with camera angles.

Video generation still consumes credits, but the core WorldGen workflow is not charged per click, making it more practical for heavy experimentation.

Exporting to Blender, Unity, and Unreal Engine

One of WorldGen’s biggest strengths is how easily you can move your scenes into popular 3D tools and game engines.

Blender export with splats support

WorldGen lets you export a Blender project that includes both the 3D objects and the 3DGS environment. Blender doesn’t natively support splats, so Rodin provides a custom script inside the Blender project. When you run it, the script sets up proxy objects and handles the splats rendering for you.

The result is a surprisingly smooth experience: you get high-quality splat backgrounds directly in Blender, alongside standard meshes for your objects. From there, you can tweak materials, adjust lighting, and integrate the scene into your usual workflow.

Mesh exports and topology options

For the 3D objects themselves, you can export in several formats and shading setups, including PBR-friendly options. If you need cleaner topology for further editing, retopology, or animation, you can use refine options during generation to get more structured meshes.

Using WorldGen scenes in Unreal Engine

WorldGen scenes can also be brought into Unreal Engine. With plugins like NanoJS for handling 3DGS, you can load the splat-based environments and combine them with the exported meshes.

Because the objects are standard editable meshes and the splats are optimized, you can quickly get a full 3D location running in Unreal. This opens up possibilities for rapid level blocking, interactive prototypes, and even early-stage game environments. For a broader look at how AI is reshaping 3D workflows for architects and designers, you can also check out this deep dive into image-to-3D tools for architecture.

Practical use cases for WorldGen

WorldGen is still early, but it’s already useful across several workflows:

• Game development: Quickly extract props and environment pieces from concept art or reference photos, then export meshes and 3DGS scenes for blocking or prototyping. Even just batch-extracting objects from a single image can speed up asset creation.

• Previs and cinematics: Build rough but visually rich environments for shot planning. You can position cameras, move objects, and then use AI Render to generate stills or video for storyboards and animatics.

• Virtual production and storytelling: Create locations for narrative projects, short films, or social content. A single image can become a reusable 3D location you revisit from multiple angles.

• Robotics and simulation: With SimReady and USDZ export, you can generate physics-aware scenes for training robots or testing agents in virtual environments.

While 3DGS in games is still in its early days, the ability to mix editable meshes with splat-based backgrounds makes WorldGen a powerful tool for experimentation and hybrid pipelines.

Why this kind of tool matters

WorldGen reflects a bigger shift in 3D and game creation: moving from painstaking manual modeling toward AI-assisted scene generation. Instead of starting from a blank level editor, you can begin from a single image and quickly get to something that looks and feels like a real place.

As quality improves and workflows mature, tools like WorldGen could become a standard part of how game developers, filmmakers, and 3D artists block out ideas, build environments, and iterate on visual concepts at high speed.

If you work with 3D in any capacity—games, VFX, previs, or robotics—this is a space worth watching closely.