Changelog

Added panoramic image viewer support and options to switch between panoramic and flat viewing.

Added prompt-template support to add multiple objects without custom instructions: "Add /path/to/object1.glb, /path/to/object2.glb, /path/to/object3.glb to the scene".

Added ability to generate 3D models with polycounts lower than 40k.

Added additional panoramic image downloads for generated Gaussian splats.

Updated help menu user tour content, descriptions, and screenshots.

Improved scene editing performance by reducing damping.

Asset generation now returns more detailed error information (for example content restrictions).

Improved 3D model generation for vegetation text prompts and image-based inputs.

Reduced polycount of generated floor terrains and complex walls to 10k.

Fixed issues when a referenced style guide is not actually a file (for example missing style guides and related cases).

Added multi-file selection workflows including delete all, drag-and-drop to chat, add to scene file, and move files.

Added automatic installation and upgrading of the Blender add-on for auto-downloaded Blender instances.

Added ability to edit scene.glb in Blender directly via right-click context menu. Updates can be saved using the plugin button "Send Scene to WorldEngen".

Added loading text for the thinking process.

Added links that land on specific sections of the Help & Learn page.

Chat history is now stored within each project folder.

Added support for creating a floor from an image. Prompt template: "Create a floor from <path to image>".

Added support for creating walls from an image. Prompt template: "Create a wall from <path to image>".

Added more detailed reporting of tasks running on the local machine.

Improved 3D model generation quality and reliability across image, multi-view image, and text workflows.

The app now prevents closing while there is an active generation in notifications.

Added quick media-asset preview support while viewing scene.glb.

Added a help menu and help guide system.

Notification panel now closes when users click away.

Added per-file camera pose persistence for SPZ files.

Added a flip orientation button in the SPZ viewer.

Dropped GLB objects are added at the current cursor position in edit mode.

Dropping a GLB in view mode switches to edit mode and adds the object to the scene center.

Added a right-click option to select polygon count before sending a 3D model prompt to chat.

Added support to generate panoramic Gaussian splats from text prompts and images, with viewer controls for rotation and movement.

Panoramic image generation now supports using another image as input (approximately 5 minutes runtime).

Added Discard Changes and Undo support for scene files in edit mode.

Skybox generation now uses the configured default style guide (custom style guide selection for this flow is not yet available).

Auto-generated file names are now capped at 3 words instead of 4.

Fixed default style guide behavior to auto-set Style Guide.txt when starting a new project.

Fixed an erroneous error message displayed on app startup.

Right-click context menu in scene files now closes when users click elsewhere.

Fixed errors related to failed auto top-up payments.

Users can now choose a location and enter a project name directly in the app.

Users can open an existing project by selecting it through Windows File Explorer.

Links from chat responses now open in the default external browser.

Added accessibility buttons in both scene view and edit mode for scene.glb.

Removed auto-save behavior for scene.glb.

Implemented manual save for scene.glb and prevents tab closure when unsaved changes exist.

Added ability to detect unsupported actions and respond with guidance toward supported workflows related to the user request.

Added support for multi-view image inputs. Prompt template: "Create a 3D model. Front view: /path/to/front.png, Back view: /path/to/back.png, Left view: /path/to/left.png, Right view: /path/to/right.png". Requires a front image plus at least one additional viewpoint. Viewpoints can also be inferred from descriptive filenames (for example image_front.png).

Added ability to create a new image from a text prompt while using another image as a reference.

Added ability to restyle an image by using another image as a reference.

Version 0.3.3 only included minor logging updates and was otherwise functionally equivalent to 0.3.2.

Improved differentiation between unsupported full-scene creation vs supported 3D-model creation, and between image-restyle vs image-editing requests when styling instructions are included.

Extended timeout for downloading 3D models from 5 minutes to 20 minutes.

Adjusted style-guide generation from images to use descriptive color names rather than hex values.

Added file extension checks for style-guide image workflows to prevent failures when users provide an image as a style guide.

Added an expired grants scanner job to trigger auto top-up for expiring trial credits.

Added new project window functionality.

Added support for opening URLs in external applications.

Added the ability for the system to provide guidance when users attempt unsupported actions.

Added a faster “add object to scene” workflow when no custom instructions are provided. Trigger via prompt template: "Add @path/to/model.glb to the scene".

Implemented a manual edit mode for scene.glb that matches the previous 3D model editor behavior. Must be explicitly enabled via the “Edit” button in the 3D viewer. Note: All other 3D models currently support view-only mode only.

Added the ability to edit an image using another image as input. Trigger via prompt template: "Edit image @path/to/image.png with the following edits: (custom instructions) using @path/to/another/image.png". Best suited for embedding a specific object from one image into another (for example placing a specific sword into a character’s hands).

General system updates to improve error handling, result parsing, tracing, and overall stability.

Updated default polycount to 100k.

Implemented a default view-only mode for scene.glb. Transforms cannot be edited in view-only mode. The P key still toggles object labels for clarity.

Rendering optimized for improved memory usage and runtime performance.

Version 0.3.1 was functionally the same as 0.3.2, except it was missing a fix for local server timeouts that could occur when Blender took too long to download. This fix is included in 0.3.2.

Removed axis flickering in the 3D viewer.

Added a background plugin status checker with UI indicators in the system panel.

Added plugin status validation before sending assets or scenes via the right-click menu.

If a plugin is not enabled and a user attempts to send an asset or scene, a UI pop-up is shown with brief instructions on how to get and install the required plugin.

Updated default Blender version from 4.5.3 to 5.0.0, with automated cleanup of the old Blender version after upgrade.

Added support for custom instructions in image-kit creation. Custom instructions are limited to image-kit operations and do not support non-image-kit file-management requests.

Updated systems so workflows with multiple activities (for example upload/download) stay on the same originating user machine.

Improved handling and logging for scene editor actions (add/remove/replace/modify objects, apply textures to floor/wall) and 3D greybox scene generation from layout flows, with more detailed results and error reporting.

Added variable-resolution textures to optimized-model workflows, intended to fix errors where 3D models could not be optimized.

This release is a breaking change and users are required to update.

Automatically sets the default style guide when a new project is created.

Added thumbs up/down controls with a feedback dialog box.

Image-to-3D and text-to-3D generation using Tencent models are now first-class, stable features. No more experimental flag required. This is the best performing model worldwide and we are the only ones who offer it in North America.

Clear visual signals help you instantly understand what task the agent kicked off.

You can now approve or ignore outputs directly through chat. (Still early, especially with long filepaths in message history.)

Ensures consistent visual direction across all generated images with zero extra steps.

Now focuses on extracting the signature vibe of your reference rather than literal objects or characters. Better for production style consistency.

Improves stability when restyling images that previously caused errors.

Improves task stability during rapid file or project switching.

One of the most requested features. You can now instantly search through previous conversations for commands, prompts, or output details.

Previously hidden behind an experimental toggle, the slider is now in the main Settings menu so it’s always accessible.

Prevents the “error aborted” messages caused by sending too early. Major quality-of-life win.

No more auto-send — prepares for upcoming custom instructions and gives users more control over generation.

Reorganizing files no longer triggers unwanted folder auto-opening.

Log out from the web UI and back in without leaving or refreshing the page.

Reduces flaky behavior during project changes and rapid switching.

Added clearer visual feedback during launch for smoother onboarding.

You can now create a style guide directly from an image or image description. Example: "Create a style guide from @path/to/image.png."

WorldEngen can now reference a specific style guide during text-to-image generation. If none is specified, the system automatically selects the largest text file as the default style guide.

Added support for converting between multiple 3D file formats (.glb, .fbx, .obj, .usd, .usdz, .usda, .usdc, .ply, .stl, .blend). Note: Some data loss may occur when converting between incompatible file types (for example, .glb → .stl).

Tabs now support temporary file previews (similar to VS Code). Single-click a file to open it in preview mode; preview tabs display in italics.

Adjust the chat box height to suit your workspace.

Zoom in and out of images for better inspection.

Added a new "Convert To" option in the right-click menu to easily convert models between GLB, FBX, OBJ, and USDZ formats.

A short video was added to demonstrate how to remove floors for AI generated 3D models.

A short video was added to demonstrate how to remove floors for AI generated 3D models.

A new helper agent was added to WorldEngen that helps you work on projects by answering your questions.

WorldEngen connects to only one instance of each app (Blender, Unity, Unreal) if multiple versions are open.