Lux3d

---

name: lux3d description: "Use Lux3D to generate 3D models from images or text. Trigger when the user asks for image to 3D, text to 3D, prompt to 3D, create a 3D model from a description, or create a 3D model from a description plus a reference image."

---

What This Skill Does

Lux3D generates 3D assets through two documented asynchronous workflows:

Image to 3D: submit an input image, poll the task, then download the ZIP result.

Text to 3D: submit a prompt plus style, optionally with a reference image, poll the task, then download the ZIP result.

Both workflows require LUX3D_API_KEY, which is a base64 invitation code in the documented format:

version:appkey:appsecret:appuid

Setup

Apply for an API key:

Form: https://forms.cloud.microsoft/pages/responsepage.aspx?id=DQSIkWdsW0yxEjajBLZtrQAAAAAAAAAAAAN__6yDykBUMDFaQzhKS1VLQlQ1UktQUU5TQjhYVkZTMS4u&route=shorturl

Contact: lux3d@qunhemail.com

Set the environment variable before use:

export LUX3D_API_KEY="your_base64_invitation_code"

Optional API root override:

export LUX3D_BASE_URL="https://api.coohom.com/global"

Documented API roots:

International: https://api.coohom.com/global

Domestic: https://api.kujiale.com/p/openapi/v2

Python Usage

Image to 3D:

from lux3d_client import generate_3d_model
result = generate_3d_model("path/to/input.jpg")
print(result)

Text to 3D:

from lux3d_client import generate_text_to_3d
result = generate_text_to_3d(
    "Generate a high-quality 3D wooden chair",
    style="photorealistic",
)
print(result)

Text plus reference image:

from lux3d_client import generate_text_to_3d
result = generate_text_to_3d(
    "Generate a premium ceramic vase with a glossy glaze",
    style="glass",
    image_path="path/to/reference.png",
)
print(result)

Low-level task APIs:

from lux3d_client import (
    create_task,
    create_text_to_3d_task,
    query_task_status,
    download_model,
)
image_task_id = create_task("path/to/input.jpg")
text_task_id = create_text_to_3d_task(
    "Generate a stylized toy robot",
    style="cartoon",
    image_path="path/to/reference.png",
)
image_model_url = query_task_status(image_task_id)
text_model_url = query_task_status(text_task_id)
download_model(image_model_url, "image_to_3d.zip")
download_model(text_model_url, "text_to_3d.zip")

Command Line Usage

Historical image-to-3D form:

python lux3d_client.py input.jpg output.zip

Explicit image-to-3D command:

python lux3d_client.py image input.jpg output.zip

Text-to-3D command:

python lux3d_client.py text "Generate a high-quality 3D wooden chair" output.zip --style photorealistic

Text-to-3D with a reference image:

python lux3d_client.py text "Generate a futuristic desk lamp" output.zip --style cyberpunk --image ref.png

Text-to-3D Styles

Supported documented styles:

photorealistic

cartoon

anime

hand_painted

cyberpunk

fantasy

glass

Output

The result URL typically points to a ZIP file. The package commonly contains:

a GLB model file

PBR texture assets

The documented result URL validity window is 2 hours.

Notes

The skill uses the documented signed API flow:

sign = MD5(appsecret + appkey + appuid + timestamp)

Image-to-3D and text-to-3D use different create endpoints.

Both workflows share the same task query endpoint.

prompt and style are required for text-to-3D.

img is optional for text-to-3D and should be a full data URL after encoding.

Requirements

pip install Pillow requests

References

ClawHub: https://clawhub.ai/violalulu/lux3d

ComfyUI reference implementation: https://github.com/manycore-research/ComfyUI-Lux3D

API contact: lux3d@qunhemail.com

Apply for an API key:

Form: https://forms.cloud.microsoft/pages/responsepage.aspx?id=DQSIkWdsW0yxEjajBLZtrQAAAAAAAAAAAAN__6yDykBUMDFaQzhKS1VLQlQ1UktQUU5TQjhYVkZTMS4u&route=shorturl

Contact: lux3d@qunhemail.com

Set the environment variable before use:

export LUX3D_API_KEY="your_base64_invitation_code"

Optional API root override:

export LUX3D_BASE_URL="https://api.coohom.com/global"

Documented API roots:

International: https://api.coohom.com/global

Domestic: https://api.kujiale.com/p/openapi/v2

The skill uses the documented signed API flow:

sign = MD5(appsecret + appkey + appuid + timestamp)

Image-to-3D and text-to-3D use different create endpoints.

Both workflows share the same task query endpoint.

prompt and style are required for text-to-3D.

img is optional for text-to-3D and should be a full data URL after encoding.