> ## Documentation Index
> Fetch the complete documentation index at: https://developers.eyequant.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Predictions Model: Attention, Clarity, and Excitingness

> EyeQuant supports three prediction types — attention, clarity, and excitingness. Learn what each predicts and what outputs are available.

The `predictions` field gives you detailed control over what is calculated in an analysis and how results are presented to you. If you omit it entirely, EyeQuant runs an attention analysis with its default outputs. When you include it, you can select one or more prediction types and, optionally, specify exactly which outputs you want back.

## Available Predictions

EyeQuant currently supports three prediction types:

| Prediction     | Description                                    | Default outputs |
| -------------- | ---------------------------------------------- | --------------- |
| `attention`    | Which elements attract viewer attention?       | `attentionMap`  |
| `clarity`      | How clear or cluttered is the image perceived? | `score`         |
| `excitingness` | How exciting is the image perceived?           | `score`         |

## Attention

The attention prediction models which parts of the image are most likely to capture a viewer's gaze. It produces up to three visual outputs:

| Output          | Description                                                                              |
| --------------- | ---------------------------------------------------------------------------------------- |
| `attentionMap`  | Heatmap showing predicted fixation density across the image.                             |
| `perceptionMap` | Visualises which areas immediately attract a viewer's attention.                         |
| `hotspotsMap`   | Circles mark the most attention-grabbing spots — larger circles indicate more attention. |

## Clarity

The clarity prediction measures how clear or cluttered your image is perceived to be. It produces the following outputs:

| Output  | Description                                                                                 |
| ------- | ------------------------------------------------------------------------------------------- |
| `score` | 0–100 score quantifying the clarity rating.                                                 |
| `map`   | Highlights how individual areas contribute to the overall perception of clarity or clutter. |

## Excitingness

The excitingness prediction estimates how exciting your image is perceived to be. It produces one output:

| Output  | Description                                         |
| ------- | --------------------------------------------------- |
| `score` | 0–100 score quantifying the predicted excitingness. |

## Configuring Predictions

You configure predictions by passing a `predictions` object where each key is a prediction name and its value is a configuration object. To use a prediction's default outputs, pass an empty object `{}`. To request specific outputs, include an `outputs` array.

**Default attention analysis (returns `attentionMap` only):**

```json theme={null}
{
  "attention": {}
}
```

**Multiple predictions with specific outputs — clarity with both score and map, plus excitingness with its default:**

```json theme={null}
{
  "clarity": {
    "outputs": ["score", "map"]
  },
  "excitingness": {}
}
```

<Warning>
  Currently, all available outputs are returned regardless of which specific outputs you request in the `outputs` array. This behavior may change to returning only the requested outputs without notice in a future version. Build your integration to handle all possible output fields defensively.
</Warning>