How does the API handle mesh and texture generation?

The generation process is split into two phases: a base mesh (without textures) is generated in approximately 90 seconds, followed by separate PBR texture map generation. The total time for a complete textured model is just over 2 minutes.

What programming languages and workflows are supported?

The API can be integrated with any platform using standard HTTP requests. We provide Python and JavaScript integration guides, making it simple to connect with AI agents, automation workflows, and custom graphics pipeline toolchains.

How to Integrate Neural4D API into Custom Workflows

Q: What is the Neural4D 3D generation API?

The Neural4D API allows developers to generate watertight 3D models programmatically. It is powered by the Direct3D-S2 engine, utilizing Spatial Sparse Attention (SSA) for high-performance generation up to 2048³ resolution.

What is the Neural4D 3D generation API?

The Neural4D API provides developers with programmatic access to the Direct3D-S2 engine (NeurIPS 2025). This engine generates watertight meshes directly from text or image prompts, bypassing traditional polygon sculpting workflows. The core algorithm employs a Spatial Sparse Attention (SSA) mechanism to achieve high geometric fidelity, producing clean quad-dominant meshes up to 2048³ voxel resolution.

How to connect via REST API (Python Example)

Neural4D API is a language-agnostic REST interface, meaning it can be integrated into any automated software applications using HTTP requests (Node.js, C#, Python, etc.). Before starting, ensure you get your Neural4D API Key. The API separates base mesh generation from texture map calculation. Base mesh creation takes approximately 90 seconds. You can explore how to implement this in a Python 3D model generator API or a generic Text to 3D API integration. Below is an integration example using Python:

python

import requests
import time

API_KEY = "your_api_key_here"
GENERATE_URL = "https://alb.neural4d.com:3000/api/generateModelWithText"
RETRIEVE_URL = "https://alb.neural4d.com:3000/api/retrieveModel"

headers = {
    "Authorization": f"Bearer {API_KEY}",
    "Content-Type": "application/json;charset=utf-8"
}

# Generate a watertight mesh with clean topology
payload = {
    "prompt": "mechanical gear, industrial grade, quad-dominant",
    "modelCount": 1,
    "disablePbr": 0
}

response = requests.post(GENERATE_URL, json=payload, headers=headers)
data = response.json()
uuids = data.get("uuids", [])

if not uuids:
    print("Failed to start generation.")
else:
    target_uuid = uuids[0]
    print(f"Task created with UUID: {target_uuid}. Processing mesh generation...")
    
    # Polling for task status (Base mesh: ~90s, Textures: ~30s additional)
    while True:
        status_resp = requests.post(RETRIEVE_URL, json={"uuid": target_uuid}, headers=headers)
        result = status_resp.json()
        code_status = result.get("codeStatus")
        
        if code_status == 0:
            print(f"Generation successful. Download GLB at: {result.get('modelUrl')}")
            break
        elif code_status in [-1, -2, -3]:
            print(f"Generation failed. Status Code: {code_status}")
            break
            
        time.sleep(10)

How to extend the pipeline to image and video generation?

The same Bearer token also unlocks media generation endpoints. Picture and video tasks accept a text prompt with up to six reference images, submit one to four jobs per request, and use multipart/form-data for uploads instead of application/json. You poll the result with the queryGenerationResult endpoint, which returns a download URL once the task reaches the completed status.

Endpoint	Purpose	Key Parameters
/api/createNormalPicture	Generate images from a text prompt with optional reference images	prompt, images (up to 6), modelKey (default image-2), aspectRatio (1:1, 16:9, 9:16, 4:3, 3:4), jobNum (1 to 4)
/api/createNormalVideo	Generate videos from a text prompt with optional reference images	prompt, images (up to 6), aspectRatio (default 16:9), duration (4 to 15 s, default 5), resolution (480p, 720p, 1080p, default 720p), jobNum (1 to 4)
/api/queryGenerationResult	Poll the result of a picture or video task by UUID	uuid; returns data.status (completed, queued, processing, failed), data.resultType (image, video), data.resultUrl

Poll queryGenerationResult with the UUID returned by the create endpoint until data.status becomes completed, then download the asset from data.resultUrl. Unlike generation, this query does not consume points.

What query parameters optimize mesh and texture outputs?

Use the query parameters below to customize mesh and texture generation outputs. For advanced configurations and endpoint details, please refer to the official generateModelWithText API reference.

Parameter	Type	Description & Range
prompt	String	The description of the 3D model. Be specific about geometric features (e.g., watertight, quad-dominant).
modelCount	Integer	The number of models planned for generation. The default value is 4.
disablePbr	Integer	PBR feature toggle. 0 = PBR enabled, 1 = PBR disabled. The default value is 0.
onlyGenerateMesh	Boolean	Generate a base mesh only, without PBR textures. The default value is false.
mesh_quality	String	Mesh density target: standard, high, or extra_high. The default value is high.
faceNum	Integer	Target face count. For standard quality, 100,000 to 500,000 (default 500,000). For high and extra_high, 500,000 to 1,000,000 (default 1,000,000).

What industrial and developer workflows benefit from the API?

Integrating the 3D generation API into developer pipelines supports multiple industries. You can learn more about its impact in our article on why your next project needs an Image to 3D API:

Game Dev Pipelines: Automated asset draft generation directly populates scene containers during runtime or editor sessions.
Industrial Prototyping: Watertight mesh generation outputs can be sent to slicing software for immediate 3D print testing.
WebAR Applications: Fast delivery of optimized GLB files enables real-time 3D asset preview on mobile browsers.

Neural4D API Integration Workflow Diagram showing requests flow from Python client code to cloud Direct3D-S2 engine and returning watertight 3D models

Need further assistance? If you have any questions or encounter issues during integration, you can provide feedback online at Neural4D Feedback or contact our technical team via email at support@neural4d.com.

How to Integrate Neural4D API into Custom Workflows

What is the Neural4D 3D generation API?

How to connect via REST API (Python Example)

How to extend the pipeline to image and video generation?

What query parameters optimize mesh and texture outputs?

What industrial and developer workflows benefit from the API?

PRODUCTS

FEATURES

COMPANY

RESOURCES