Nova3D is an AI 3D generation tool that produces structured, editable, part-aware 3D assets by generating Blender-native construction scripts and exporting a named, hierarchical GLB instead of a single merged mesh.
https://github.com/RareSense/Nova3D?ref=producthunt
Nova3D

Product Information

Updated:May 18, 2026

What is Nova3D

Nova3D is a “prompt-to-code” (and image-to-code) 3D asset generator that targets Blender as its native representation, allowing it to output models as executable Blender Python construction procedures rather than mesh-only results. The system emphasizes preserving part identity, hierarchy, transforms, materials, and provenance, so creators can edit individual components after generation. The open-source repository provides the client UI (built with Flutter) that connects to the hosted Nova3D service (currently closed-source), where users authenticate, supply an API key for an LLM provider (OpenAI, Anthropic, or Gemini), and generate assets that can be viewed and edited in an integrated web-based viewport.

Key Features of Nova3D

Nova3D is an AI 3D generation system that produces structured, editable 3D assets by generating Blender-native Python construction scripts (prompt-to-code / image-to-code) rather than a single merged mesh. By targeting Blender’s scene graph, it preserves part identity, hierarchy, transforms, materials, and provenance, then outputs a structured GLB with named, separately addressable components. It includes a built-in web UI with an integrated viewport/editor, supports reference images, caches results locally in the browser, and is model-agnostic—letting users choose different LLM providers (e.g., OpenAI, Anthropic, Gemini) via an API key—while the open-source repo provides the client that connects to a hosted (currently closed-source) backend service.
Blender-native construction code output: Generates executable Blender Python scripts as the core artifact, enabling precise, procedural creation and regeneration of parts instead of producing an unstructured mesh blob.
Part-aware, hierarchical assets: Outputs structured GLB files with named, separately addressable parts, correct parenting/hierarchy, and preserved transforms—supporting surgical edits (e.g., modify a handle without regenerating the whole object).
Materials and PBR-friendly mapping: Maintains material assignments and supports proper PBR-style workflows rather than relying on baked vertex colors typical of many mesh-extraction pipelines.
Model-agnostic LLM harness: Works with multiple LLM providers (e.g., Claude/GPT/Gemini) configured via user API keys, while the pipeline handles validation/execution of generated Blender code.
Integrated web viewport & editing tools: Includes a built-in Three.js-based viewer/editor with transform tools, snapping, and material editing for quick iteration and adjustment.
Reference images + local caching: Allows attaching images to guide spatial/structural generation and caches models in-browser so users can access history even if remote URLs expire.

Use Cases of Nova3D

Game/film asset prototyping: Rapidly generate structured props with editable sub-parts (e.g., weapons, furniture) so artists can tweak components and materials without starting over.
Product design concept iteration: Create early-stage product forms (appliances, consumer gadgets) as editable assemblies, enabling quick part-level exploration of handles, housings, and interfaces.
E-commerce visualization & configurators: Generate modular 3D product models (separate materials/parts) that can be adjusted for variants (colors, components) for web-based previews and marketing renders.
Jewelry and small luxury CAD exploration: Produce structured, Blender-based procedural models suitable for iterative styling and component edits, aligning with the stated engine lineage used for specialized jewelry CAD.
3D printing preparation workflows: Generate models with separable parts (e.g., snap-fit assemblies or multipart figurines) that can be individually adjusted before export to printing-oriented pipelines.

Pros

Produces editable, part-aware assets (hierarchy, names, transforms) rather than a single merged mesh, enabling targeted revisions.
Leverages Blender’s full scene graph and modifiers, supporting both precision and organic forms beyond typical CSG-only approaches.
Model-agnostic setup: users can choose among multiple LLM providers using their own API keys.
Open-source client with a quick local setup and an integrated web-based viewport/editor.

Cons

Backend service is (currently) closed-source and hosted by default, so full self-hosting requires a custom backend endpoint.
Requires Flutter (and web-server/OAuth setup constraints like using 127.0.0.1) which may add friction for some users.
Quality and reliability depend on LLM-generated code correctness and the chosen model/provider behavior.

How to Use Nova3D

1) Install prerequisites: Install Flutter 3.24+ (https://flutter.dev/) on your machine so you can run the Nova3D client locally.
2) Clone the Nova3D client repository: Run: git clone https://github.com/RareSense/Nova3D.git
3) Enter the project directory: Run: cd Nova3D
4) Install dependencies: Run: flutter pub get
5) Start the local web UI (required port): Run: flutter run -d web-server --web-hostname 0.0.0.0 --web-port 5555 (Port 5555 is required for OAuth redirect authorization.)
6) Open the app in your browser: Navigate to: http://127.0.0.1:5555 (Use 127.0.0.1, not localhost, to avoid Google Sign-In OAuth origin issues.)
7) Sign in: In the Nova3D UI, sign in using Google or Email as prompted.
8) Add an LLM API key: Go to Settings → add your API key (supported options listed in the README: OpenAI, Anthropic, or Gemini).
9) Create a generation prompt: Enter a text prompt describing the 3D asset you want to generate (Nova3D generates Blender-native Python scripts and returns a structured GLB with named, separately addressable parts).
10) (Optional) Attach a reference image: If you want the generation guided by an image, use the Reference Images feature to attach a photo to guide spatial logic.
11) Generate the model: Submit the prompt to generate. The client communicates with the hosted nova3d.xyz API by default.
12) Inspect and edit in the integrated viewport: Use the built-in Three.js editor/viewport to view the GLB and make edits such as transforms, snapping, and material editing.
13) Review cached history (optional): Use the app’s local caching/history to revisit previously generated models in-browser, even if remote URLs expire.
14) Build for static hosting (optional): Create a production web build with: flutter build web --release
15) Point the client to a custom backend (optional): If you self-host a compatible backend, run the client with: --dart-define=API_BASE_URL=https://your-api.com (otherwise it uses nova3d.xyz by default).
16) Troubleshoot sign-in loops (if needed): If authentication loops occur, ensure you are using http://127.0.0.1:5555 (not localhost:5555), per the README troubleshooting note.

Nova3D FAQs

Nova3D is an AI 3D generation system that produces 3D assets as executable Blender construction procedures (Blender Python scripts) and returns a structured GLB with named, separately addressable parts, rather than a single merged mesh.

Latest AI Tools Similar to Nova3D

JustAHuman
JustAHuman
JustAHuman is a gaming platform that rewards players for completing challenges while helping game creators process 3D assets through AI.
Sketcho
Sketcho
Sketcho is an AI-powered design tool that transforms sketches and ideas into high-quality professional designs through an intuitive interface.
Rendair
Rendair
Rendair is an all-in-one AI-powered architectural rendering platform that offers quick, high-quality visualizations through both AI tools and professional 3D artists for architects, designers, and real estate professionals.
Triorama AI
Triorama AI
Triorama AI is an AI-powered 3D product configurator platform that enables eCommerce businesses to offer real-time product personalization and visualization capabilities to their customers.