AI art has exploded onto the scene in the past few months with the rise of some powerful AI image generators. With this sudden rise of these tools, an overwhelming number of people started to use the platform, ultimately creating a buzz within the AI fine art scene. 

To get better results with AI image generators, learning how to write good prompts is extremely important. AI writing generators like ChatGPT return decent responses with simple prompts, but prompts for AI image generators require more attention to detail. This guide will explain how to write effective prompts for AI image generators to get exactly what you want.

Simple AI Art Prompt Structure

Instructions: Try combining these elements to create your own AI art prompt! (separate each element by commas)

1. Subject

The main content of your choice

2. Medium

painting, sculpture, watercolor, oil painting, drawing

3. Style

pop art, impressionism, modern art, full list

4. Artist

Pablo Picasso, Salvador Dali, Steve Ditko

5. Resolution

ultra-detailed, sharp focus, photorealistic

6. Lighting

soft light, hard light, dark, cinematic

7. Color

ocean blue, neon pink, lime green, full list

8. Additional Details

Any missing factors

Before We Start

Unlike AI writing generators, you must be as descriptive as possible when generating AI images. Generative AI tools are not good with subjective and broad prompts. For instance, a “cute dog” would not be sufficient to render the image that you may be looking for. These tools only generate images they are trained on, so writing details about features and patterns is much more helpful for the AI image generator. Ready to explore AI image generators now? Check out my top picks on AI image generators here.

Is there a limit to AI art prompts?

This depends on which platform you use. For instance, Midjourney operates well with long prompts over 50 words, but some tools like Stable Diffusion work best with prompts under 400 characters. As different tools have various optimal settings, it is important to try and see what works best with them.

Writing AI Art Prompts

For more simple content, the first thing that should be at the prompt is the content description. Is it a photograph we are looking for? Or is it more like an illustration? You can add this to the subject of the content:

[An illustration of + subject] 

Adding an action, an adjective, and/or an adverb may lead to a more predictable result. Remember, the more descriptive your prompt is, the more the result will be what you seek. At the same time, this may sound a bit contradicting, but it is also important not to overwhelm the tool with long sentences; for more complex content, it is important to separate different features (style, artist references) with commas. For more intricate art, it would be nice to organize it in the following order.

[Subject, medium, style, artist reference, resolution, lighting, color]

Some of these features may be less intuitive, so I will review the key concepts. Medium refers to the material used to create the artwork. Some examples would be oil painting, watercolor, and sculpture. Next, we move on to style. Style describes the way the artwork looks. Pop art, realism, and modern art are some major examples. Tools like Midjourney work well when you add artist references; it enables you to mimic the style of the artist of your choice. Regarding colors, AI art generators cannot recognize HEX or RGB color codes, so you would need to know exactly what colors you want in your image. Color extraction tools such as TinEye will tell you which color is included in the image. Again, coming up with the image you want requires a long process of trial and error; there is a table of potential factors that can be added to your prompt at the beginning of this guide.

Many of these tools have their characteristics; some of the observations for Midjourney are:

The words at the beginning of the prompt are more important.

There are advanced parameters that can be added at the end with a double hyphen (see full list)

Adding an artist reference will fine-tune the content. For Stable Diffusion, one of the biggest differentiators is the negative prompt feature, where users can write what they don’t want and what they want.

Remembering all these parameters seem a bit overwhelming? Another useful tip is to use an image-to-text generator like CLIP interrogator. Uploading an AI image will give you a text prompt, and you can discover what kind of written prompt generates the image using a reverse engineering technique. Take that prompt as an example and start building content from there.

Categories: Uncategorized