Generating images with the Web Editor – Evolv AI

Medium: Portrait, digital painting
Style: Hyperrealistic, pop-art
Artist: Any artist by name
Website: Pixabay, artstation
Resolution: ALLI generates images based on the component size
Additional details: Dramatic, psychedelic
Color: Vintage, silver
Lighting: Cinematic lighting, rim lighting

This articles describes how to use the Web Editor to create AI-generated images to use in your optimization.

Table of Contents

Steps to create AI-generated images

Open the Web Editor.
- If you're new to Evolv AI, check out how to create a context and variable for your project.
Create a component that contains the image you want to experiment with.
- The component can be a <img> tag or an HTML element with an image background.
Create a new variant.
You can name the variant using the text prompt you need to generate the required image.
- For best results, use the format: [Component Name] is "prompt"
  - E.g., Hero Image is a Steampunk Submarine
- When you name variants in the Web Editor, ALLI (the AI assistant) generates the necessary code, or in this case, the required image assets.
  - ALLI will only generate code the first time you name the variant, provided no changes have been made.
- If you didn't name the variant or used an arbitrary name instead, open the chat panel and enter your prompt for ALLI to generate the image
Three generated images will appear in the chat panel.
- In this example, the prompt has been expanded from the original variant name to include more specific details. See How to craft an effective prompt.
Hover the mouse over the desired image and click APPLY to select it.
- Applying an image selects it for use in the current variant.
- You can apply different images to see how they look on the page. The variant remembers the last applied image.
- You can also download the images or copy their URLs by clicking the three dots at the bottom-right of each image.

Prompts for AI-generated images are complex. They include different elements to tell the AI model exactly how you want your image to look.

Every prompt needs a subject (the "what"), but you can also include other descriptors. Here are some examples.