# Gen3D -> Full PBR - Workflow

### What you'll need

* A web browser
* A Github account to log into replicate
* A credit card for Replicate (pay-per-run; a single Hunyuan3D run costs cents, not dollars)
* An image or text prompt for the asset you want to create
* A Marso Studio account (with access granted) — sign up at [app.m-xr.com](https://app.m-xr.com/) if you haven't already

***

### Step 1 — Create a Replicate account

1. Go to [replicate.com](https://replicate.com/)
2. Click **Sign in** (top right) and sign up with GitHub, Google, or email
3. Once signed in, head to **Billing** in your account settings and add a payment method. Replicate charges per second of GPU time — Hunyuan3D 3.1 typically runs in under a minute per asset.

{% hint style="info" %}
When running this model internally, the cost is 16 cents per generation. However this does not count as a guarantee for your generation.
{% endhint %}

***

### Step 2 — Run Hunyuan3D 3.1

1. Go to the model page: [replicate.com/tencent/hunyuan-3d-3.1](https://replicate.com/tencent/hunyuan-3d-3.1)
2. You'll see an input panel on the left and an output preview on the right.
3. Provide your input:
   * **Image-to-3D** (recommended): upload a clean reference image of your subject — ideally a single object on a plain background. Front-facing shots work best.
   * **Text-to-3D**: enter a descriptive prompt (e.g. *"a weathered wooden treasure chest with iron banding"*).

{% hint style="success" %}
We’ve found best results have been when using an image as a starting point.
{% endhint %}

4. Leave the default settings as-is for your first run. They're tuned for quality.
5. Make sure to keep the PBR toggle **OFF**.
6. Click **Run**. The model will take 30–90 seconds depending on queue time.
7. When it finishes, you'll see a 3D preview on the right.

***

### Step 3 — Export the GLB

1. Below the 3D preview on the Replicate output panel, click the **download icon** next to the output file.
2. Save the `.glb` file somewhere you'll find it (Desktop or Downloads is fine).

That's your raw mesh. Now we texture it.

***

### Step 4 — Bring the GLB into Marso Studio

1. Go to [app.m-xr.com](https://app.m-xr.com/) and sign in.
2. Open **Marso Studio** from your dashboard.
3. Click **New project** (or drag your GLB straight onto the upload area).
4. Drop in the `.glb` you exported from Replicate.
5. Marso Studio will parse the mesh and show you a preview with the existing baked textures stripped out — this is intentional. You're about to replace them with proper PBR materials.

***

### Step 5 — Run the I2M model in Marso Studio

1. With your asset loaded, hook up a PBR generation node and click **Generate PBR**.
2. Marso Studio's I2M (Image-to-Material) model will analyse the geometry and produce a full set of PBR maps:
   * Base colour (albedo)
   * Roughness
   * Metallic
   * IOR
3. Generation typically takes 4–6 minutes.
4. Once complete, you'll see your asset re-rendered with the new materials in the preview window.

***

### Step 6 — Review and export

1. Use the viewer controls to orbit the asset and inspect the materials under different lighting.
2. If you want to iterate, you can re-run with adjusted parameters or swap reference images.
3. When you're happy, click **Export**. You'll get:
   * The textured GLB (drop straight into Unreal, Unity, Blender, Maya)
   * Individual PBR map files at your chosen resolution

{% hint style="info" %}
We are still working on the viewer within Marso Studio, it may be beneficial to export and view the materials in another platform.
{% endhint %}

***

### Where to take it next

* **Drop into Unreal**: import the GLB into your Unreal project — materials apply automatically.
* **Use in Blender**: import the GLB into Blender's scene view; the PBR maps connect to Principled BSDF out of the box.

***

### Why this workflow matters

AI mesh generation can produce good quality geometry. The remaining bottleneck is materials — most generated meshes ship with low-quality, baked textures that fall apart under production lighting and can't be edited.

M-XR's I2M model is trained on a proprietary dataset of **measured** PBR materials captured from real-world objects. That means the materials you get out aren't artist-interpreted or estimated — they're grounded in objective real-world data.

Combine the two and you've got a pipeline from text prompt to production-ready, fully textured asset in under 10 minutes.

***

### Our Results

{% embed url="<https://sketchfab.com/3d-models/auto-pbr-hunyuan-old-blue-car-22a784a06ea9408f914eb444e1ce24ae>" %}

{% embed url="<https://sketchfab.com/3d-models/auto-pbr-hunyaun-old-sword-2ce0b50f7cd140a89d4f9a2c9c427683>" %}

{% embed url="<https://sketchfab.com/3d-models/auto-pbr-hunyaun-samurai-helment-94bbb1ed3ac04aa6bd7a931839ec0e93>" %}


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.m-xr.com/marso-studio/getting-started/gen3d-greater-than-full-pbr-workflow.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
