---
title: "Image Studio"
description: "Image Studio skill for Vellum — generate and edit images using AI through your assistant."
canonical_url: "https://www.vellum.ai/docs/skills-reference/image-studio"
md_url: "https://www.vellum.ai/md/docs/skills-reference/image-studio"
related:
  - "/docs/skills-reference"
  - "/docs/skills-reference/acp"
  - "/docs/skills-reference/amazon"
  - "/docs/skills-reference/app-builder"
  - "/docs/skills-reference/browser"
  - "/docs/skills-reference/chatgpt-import"
  - "/docs/skills-reference/computer-use"
  - "/docs/skills-reference/contacts"
---

# Image Studio

## What it does

Generates and edits images using AI models. Create illustrations, modify photos, generate art, and produce visual assets from text descriptions.

## Setup required

A Gemini API key is required. If one isn't configured, your assistant will prompt you to set it up on first use

## Permissions

- No macOS permissions needed for generation
- File system permission needed if saving images to your machine

## Common prompts

| You say...                                       | What happens                                        |
| ------------------------------------------------ | --------------------------------------------------- |
| “Generate an image of a sunset over mountains”   | Creates an AI-generated image from your description |
| “Make me a logo for a coffee shop called 'Brew'” | Generates a logo design                             |
| “Edit this image to remove the background”       | Modifies an existing image                          |
| “Create a cartoon version of this photo”         | Transforms an image style                           |
| “Make a banner image for my blog post about AI”  | Generates sized-for-purpose assets                  |
| “Generate 4 variations of this design”           | Multiple options to choose from                     |

## Configuration

- **Models:** Nano Banana 2 (default, fast) or Nano Banana Pro (higher quality, slower)
- **Variants:** Generate 1–4 variations per prompt to pick the best result
- **Modes:** Text-to-image (generate from a prompt) or edit mode (modify an existing image)

## Tips & gotchas

- **Be descriptive:** The more detail in your prompt, the better the result. “A watercolor painting of a golden retriever sitting in a field of lavender at sunset” beats “a dog.”
- **Iterate:** First generation not perfect? Say “make it more colorful” or “try a different angle.” Your assistant refines based on feedback.
- **Formats:** Images are generated as PNG or JPEG. Specify if you have a preference.
- **File delivery:** Generated images are delivered as attachments in chat. You can also ask your assistant to save them to a specific folder on your machine (requires file access permission).