TL;DR

Extract UI Elements fits a concrete game-scene screenshot when the goal is to pull UI pieces into one sprite bitmap via a vision model. It is a poor fit when region count and boundaries must be exact: results stay approximate and depend on the chosen model.

Extract UI Elements runs model-based UI extraction on the current image. No prompt is required; the flow is driven from the UI.

Step 1: Select the feature from the menu

With the target image selected, open the Extract-related menu and choose Extract UI Elements.

Toolbar Extract menu with Extract UI Elements highlighted

Step 2: Set the model and parameters

In the dialog, pick the model and output size. Supported options include GPT Image 2, Nano Banana / Nano Banana 2, and Nano Banana Pro. Larger sizes tend to yield more items and more variety in the extraction.

Extract UI Elements dialog with model and output size options

Step 3: Check the extraction result

After generation, review the assembled sprite. The current output is a non-transparent bitmap layer; cleanup is usually needed before feeding it into later steps.

Extracted UI elements assembled on a non-transparent sprite bitmap
Result with GPT Image 2, Output Size 1K.

Better fit: bitmap to sprite sheet

A UI-dense single bitmap is a stronger scene for Extract UI Elements than the walkthrough above. The model pulls discrete controls and panels from one source image into a single sprite map; coverage and boundaries here are richer and more precise than the Step 3 demo.

UI-dense game bitmap used as the Extract UI Elements source
Source: one UI-dense bitmap.
Extracted UI elements laid out on a single sprite sheet
Output sprite sheet with GPT Image 2, Output Size 1K.