# Prepare a model

URL: https://docs.glossi.io/bring-in-your-product/prepare-a-model

Split a one-mesh model into parts, group parts by finish, redo UVs, assign materials or match a photo, then keep the result or try again.

Model prep looks at your model the way a 3D artist would and gets it ready for materials and the Studio. It runs on a model already in your library and never changes it until you keep the result.

## Start a prep

1. Open the model in the library and go to the **Actions** tab.
2. Choose an action:
   * **Prepare model** runs everything below in one go. On a model that's one mesh it splits it first; on a model that already has parts it groups them instead.
   * **Split into parts** separates a model that arrived as a single mesh, common with scans and generated models, into its real parts, such as the upper, sole and laces of a shoe. A sewn garment splits along its pattern seams. The model looks the same afterwards; only its parts change. Models exported from a 3D or CAD tool already have parts and don't need this.
   * **Group parts by finish** names every part and groups the parts that share a finish, so each finish gets its own material slot. Where one mesh holds separately modelled pieces in different finishes, such as a shoe's overlays and mesh panels, it pulls them apart first.
   * **UV unwrap** gives every part even, real-world-scale UVs, so a texture is the same size on every part.
   * **Assign materials** picks a material from your library or the Glossi catalog for each finish, and can make a new one when nothing fits. Parts with printed artwork keep their own look.
3. Optionally add a note the agent should know, such as "the strap and buckle are separate parts", and up to six photos of the real product. Photos help most with colors, materials and where one part ends and the next begins.
4. Select **Start**. For a paid prep, the button shows the price: **Pay $… and start**.

**Group parts by finish**, **UV unwrap** and **Assign materials** usually take a few minutes. **Split into parts**, and **Prepare model** on a model that already has parts, take up to 30 minutes. A full **Prepare model** on a one-mesh model takes up to 40 minutes. You can leave the page; progress shows on the **Actions** tab when you come back.

## Match a photo

**Match a photo**, beside **Create variant** on the **Actions** tab, makes a variant that looks like a photo of the product.

1. Select **Match a photo**.
2. Add up to six photos and name the variant.
3. Start the run. It usually takes a few minutes.

The agent groups the model's parts by the colors and materials the photos show, then picks a material for each group, tinted and scaled to match. When you keep the result, the picks are saved as a variant on the regrouped model; the model's own look doesn't change. A color printed or painted inside one piece can't be separated this way: the result lists it under **Worth a look**, and it needs **Split into parts** first.

## Review the result

When the run finishes, the result waits on the **Actions** tab.

* Read the summary and anything under **Worth a look**: the parts the agent wasn't sure about.
* Select **Preview the result** to see it in the viewer, and **Show the current model** to compare.
* A split doesn't change how the model looks, so select **Show parts** while previewing. Each part gets its own color, and hovering a part shows its name.
* **Keep** makes the result the model's current version. The previous version stays in the model's [version history](https://docs.glossi.io/bring-in-your-product/model-versions.md), so you can go back.
* **Keep as a new model** saves the result as a separate model and leaves this one as it is. When the model already has variants or is in projects, this is the default, because regrouping renames the parts' material slots and variants or Studio edits made for the old groups would no longer line up.
* **Discard** throws the result away.
* **Change something** (after a split) changes only the area you describe, for example "the laces include the red strip on the tongue". Add screenshots of the area, or select **Add this view** to attach the viewer as it is; with **Show parts** on, that shows the agent exactly which part is wrong. If the agent needs more to go on, it asks a question on the model page; answer it to make the change. If it leaves the result as it was, it says why, and the result is still there to keep or discard.

## What it costs

Splitting an uploaded model into parts (**Split into parts**, or **Prepare model** on a one-mesh model) is a one-off purchase. The price shows before you start, and it's charged when the prep starts, to the saved card or however your workspace's model purchases are set up. Up to three changes before you keep it are included. A workspace owner or admin starts a paid prep.

A paid prep isn't refunded if you discard the result, so use the included changes to fix what it got wrong. A run that fails isn't charged.

Everything else (**Group parts by finish**, **UV unwrap**, **Assign materials**, **Prepare model** on a model that already has parts, and **Match a photo**) uses your workspace's usage when the result is ready to review. A run that fails uses nothing.

Models made with AI model creation are prepared at no extra cost.

## Related

* [Model prep in the API and MCP](https://docs.glossi.io/api/model-prep.md)
* [Model versions](https://docs.glossi.io/bring-in-your-product/model-versions.md)
* [Variants](https://docs.glossi.io/set-up-the-product/variants.md)