Visión

Overview

Unlike text-only models, Vision models can process images as input and combine visual understanding with natural language reasoning. Simply provide an image along with an optional prompt, and the model returns a structured response based on what it sees.

How Vision Works

Analyze an Image

1const response = await client.vision.analyze({
2  model: "nova-vision-1",
3  image: "./workspace.jpg",
4  prompt: "Describe everything visible in this workspace."
5});
6 
7console.log(response.output);

Request Properties

Vision model identifier.

Image to analyze.

Additional instructions for the model.

Example Input & Output

Prompt

Common Use Cases

• Image captioning

• Object detection

• UI and screenshot analysis

• Document understanding

• Product image analysis

• Visual question answering

Next Step

Continue to Files to learn how to upload, manage, and reference files across NovaAI APIs.