AI Money Code – Emmanuel Cobbi: +237 672754178

How to Write AI image prompts

An ultimate guide on writing AI art prompts that deliver exceptional
results every time.

It’s 2023 and you must have seen some amazing AI art being posted by me or
other persons on the internet.
AI images go beyond the current trend of people generating AI Portraits of
themselves.
Creating AI images is an art, a global new skill that was born in the late quarter
of 2022, made official in 2023 and will be here for a very long time.
There are a lot of AI tools you can create images from but the secret isn’t in the
tools but in the prompts.
Fine-tuning your prompts help AI tools better understand your input, giving you
beautiful and detailed images to use.

Here are some of the AI tools we’ll be using to create images. Some of them are
paid, others are free. If you can purchase a paid subscription please do else it’s
okay to get started on the free tools.

Here are the site links for some AI tools that can generate images from text or
other inputs:

 Clipdrop: A suite of AI-powered image manipulation tools that can

generate high-resolution realistic images, uncrop photos, transform
doodles, remove objects, and more.
AI Money Code – Emmanuel Cobbi: +237 672754178

 Firefly: An AI tool from Adobe that can create and edit images using
natural language commands and generative models.
 Midjourney: An AI tool that can generate images from text using a large-
scale neural network model trained on diverse data sources.
 Playground AI: An AI tool that can generate images from text using a
smaller and faster neural network model trained on curated data sources.
 Lexica AI: An AI tool that can generate images from text using a hybrid
approach that combines neural networks and symbolic reasoning.

 DALL-E: An AI tool from OpenAI that can create diverse and realistic
images from any text prompt using a transformer model. You can try it
out here
 DeepAI: A platform that provides various AI image generation models,
such as faces, landscapes, cartoons, and abstract art. You can explore
their tools here
 Chimera Painter: An AI tool from Google that can create realistic
creatures from simple sketches. You can use it here
 Artbreeder: An AI tool that can create and explore images using genetic
algorithms and neural networks. You can use it here

What are AI art prompts?

The text you enter into an AI art tool is called a prompt.
A prompt can be a few words or multiple sentences.
The maximum limit is 400 characters on Dall-E, and 6,000 characters on
Midjourney. Not sure about the other tools.
Prompts don't have to be long, and AI generators can still produce images with
short text prompts, or even emojis.
The shorter a prompt is, the more you leave up to interpretation. If the style isn't
mentioned, AI art generators will use their default style.
AI Money Code – Emmanuel Cobbi: +237 672754178

On some AI tools, prompts go beyond text prompts. You can also add an image
prompt, URL for reference, and parameters for aspect ratio and upscaling.

How to write effective AI art prompts:

1. Add specific details:

Mention colours, shapes, and any other details that can create a more
accurate image. Instead of man on his laptop, write “man in his fifties,
typing on a large grey laptop, at his desk.” This way, you leave less for
AI to interpret.

2. Be brief:
Adding details help, but extra-long prompts can confuse AI generators.
So stick to active voice and space out long sentences with punctuations.
And don't repeat words or refer to the same subject twice.

For example: “Cat with orange and white spots looking happy and
playful, the cat is rolling on the grass” can be shortened to, “Orange and
white spotted cat, rolling on the grass, happy, playful.”

3. Use descriptive language:

AI image generators understand descriptive terms well, so pull out a
dictionary.

Instead of writing happy, try elated, satisfied, tranquil, or exuberant to get

a specific result. Remember, synonyms don't always give the same
results—finding the right word is key.

Also, don’t use very before words, it has limited effects on the output,
and it's better to find another adjective.

4. Use an image reference:

AI Money Code – Emmanuel Cobbi: +237 672754178

Upload an image or enter its URL in your AI generator to get a similar

image. On Midjourney for example, you just need to enter the image
URL. On Dall-E, you need to use the API to access this feature.

Image prompts work as a shortcut—AI immediately understands the

subject, context, and composition without having to describe it in text.
With this feature, you can create similar outputs from reference images
quickly.

5. Reference an artist's style:

Suppose you want a Japanese animated image that looks hand-painted,
and has vivid coloring, but doesn't look like mainstream anime. Use “in
the style of Studio Ghibli” and you'd get exactly that output, minus all the
extra words.

Adding a famous artist's or photographer's style can deliver a similar

output.

6. Reference a specific movie or artwork:

Don't want to use an artist's name? Then ask AI to create in the style of a
specific artwork.
You can also enter cinematography from “X” movie to copy the aesthetic
and overall mood.

Try famous directors, too: “in the style of Wes Anderson” or “in the style
of Quentin Tarantino” to nail an elusive style.

7. Unbundle an artist's style:

The downside of copying an artist's style in the prompt? It can get too
obvious. Try using “in the style of Van Gogh.” You'll quickly see all
images look like obvious Starry Night plagiarism.

Instead, ask ChatGPT (or any AI text generator) to describe Van Gogh's
style and use these phrases in the prompt instead.

8. Decide between creative and realistic:

When AI generators are creative, its more liberal with the image,
enhancing the style and composition, sometimes at the cost of realism. If
you want something sharp and clear, use “detailed” instead of “realistic.”
AI Money Code – Emmanuel Cobbi: +237 672754178

Only use the word realistic or realism when you'd like the image to be as
close to a photograph as possible.

9. Don't mention image quality:

Using terms 4k or 8k doesn't always improve image quality, especially on
Dall-E.

You can play around with hyper-realistic details, but the image quality
remains the same. On some AI tools, however, you can click on upscale
to improve the resolution and quality, but it doesn't need to be added to
the prompt.

10.No negative prompts:

If you want a party with no balloons in the background, don't add any
balloons to your prompt. If you do, AI will likely make sure balloons are
there in the image.

Instead, get specific on what you want to see and be specific—switch

party decorations for streamers, confetti, and posters.

11.Use Outpainting:
Outpainting is a feature that allows you to generate new parts of an
existing image.

If you'd like to zoom out and extend an AI image size, upload it into an
AI tool with a blank canvas border. Then enter a text prompt to fill in the
blank portion with new elements.

12.Use photography terms:

When generating photographs, specify technical aspects.
Here's an example for a portrait photograph to get started: “Canon EOS
5D Mark IV DSLR, f/5.6 aperture, 1/125 second shutter speed, ISO
100.”

Adding terms like award-winning or photojournalism creates a more

polished vs. candid output.

13.Search for art styles:

AI Money Code – Emmanuel Cobbi: +237 672754178

Do you know what Lettrism, Vienna Secession, or Anti-Design are? AI

art generators can recognize obscure art styles, so become acquainted
with them.

14.Fix faces and fingers:

AI art generators are notorious for messing up faces and extremities.
Some of them are not good at it yet. To fix this, use photo editing
software to delete the face or mistake.

Re-upload this picture and add a new prompt for the face.
This helps AI focus on perfecting the face instead of the entire image. Ps:
This does not work for every AI tool yet.

Simple prompt style ideas you can begin with:

You may not know what how to put the prompts in a form that is clear and
specific, but never mind, here are some prompt ideas for AI Art in detail you
can choose from.

 Adjective Prompts:

Adjectives play a crucial role in writing best prompts for AI art. They
provide descriptions of emotion and technique that can help you get desired
images from AI prompt image generator.

The adjective prompts include conceptual, aesthetic, biomorphic, beautiful,

Realistic, colorful, ball bearing, biomorphic, brushwork, classic, clever,
complex and more…
AI Money Code – Emmanuel Cobbi: +237 672754178

 Art Style Prompts:

To get the style you want, the best prompts should involve style
descriptions. Here are some for you:

Modern, Pixel art, Realism, Surrealism, Abstract, Expressionism,

Abstraction, Academic, Action painting, Aesthetic, Allover painting,
Angular, Appropriation, Architecture, Artifice, Automatism, Avant-garde,
Baroque, Bauhaus, Contemporary, Cubism, Cyberpunk, Digital art, Fantasy,
Impressionism, Minimal,Ukiyo-e print and more.

You could put your prompts as follows,

1. Ukiyo-e print of two men visiting a shrine.

2. Cyberpunk style city street at night, Expressionism.
3. Action painting of a cottage near a pond, spring time.

 Prompts in the style of non-living famous artists:

To generate some images of classic style, you could get the specific image
you want by referring to the masters of the history. But to avoid and
copyright problems, you’d better choose non-living ones like Frida Kahlo,
Rene Magritte, Tamara de Lempicka who bears distinctive art styles.

1. When mentioning the style of Frida Kahlo, the best prompts for ai should
be “Portrait of a girl next to a capybara, in the style of a Frida Kahlo
painting”.
AI Money Code – Emmanuel Cobbi: +237 672754178

2. And the prompts involving the style of Rene Magritte should be “Boy
with a cowboy hat, in the style of Rene Magritte”
3. And the prompts for the style of Tamara de Lempicka would be “ a
women with red hair and mouth in the style of Tamara de Lempicka.”

 Prompts with Pop Icons:

To make fun prompts for ai art, you could choose one with the popular
cultural icon as the subject. They could be the comic heroes, movie
characters, protagonists in video game that you prefer.

You could add descriptive words like “digital art” or “cinematic lighting” or
other kind of lighting that come out of your mind.

1. For example, when you choose Pikachu as part of your prompt, you could
write the prompts like “Pikachu, cinematic, digital art”.
2. Or you choose the Iron man, the best prompts for AI art would be “Iron
man, cinematic lighting, high resolution 3D render”
3. Storm Trooper could be your choice too like “Storm trooper in the desert,
high-detail, dramatic lighting, digital art”.
AI Money Code – Emmanuel Cobbi: +237 672754178

 Portraits Prompts:

When it comes to generating AI portrait pictures, there are countless options

available to choose from. There are many portraits for you to choose
involving human and animal as well as the AI portraits which is really
creative.

You could add words for style like watercolour painting or photorealistic or
others to get the effect you want.

You could put the prompts as follows:

1. Portrait of a dog cat wearing a robe, Chinese watercolour

painting.
2. Portrait of a Sci-Fi man, peaceful, digital art.
3. Portrait of an old lady wearing a dress, photorealistic.

 Fantasy Prompts:

The beautiful fantasy scenery of all kinds is the good elements for the best ai
prompts. You could choose themes encompassing the nature, medieval,
surreal and more.
AI Money Code – Emmanuel Cobbi: +237 672754178

Besides, you could add words to your prompts like “fantasy”, “digital art”,
“epic” or “surreal”. You could put your prompts like this,

1. "Cottage overgrown with trees and grassland, high-fantasy, digital art.

2. Medieval German castle, surrounded by mountains, high fantasy, epic,
digital art.
3. Castle on top of a cliff, surreal, fantasy, digital art.

 Some Advanced stuffs:

 Aspect Ratio:
AI Money Code – Emmanuel Cobbi: +237 672754178

Change the Aspect Ratio of an image to get different sizes. Parameters are
added to the end of a prompt with a -- prefix.

For example, the parameter is “-- ar 1:1 for a square image.”

If you're creating an image for an Instagram Story, you'd use 9:16.

If you’re creating a full-screen laptop presentation, you'd use something like

16:10 or 16:9.

 Negative prompts:

Suppose you want a garden with plants, but no flowers. Simply saying no
flowers in the prompt will cause AI to add them in since it doesn't understand
negative prompts.
Use the parameter “-- no flowers to create an image with no flowers.”

 Weights:
When you have more than one subject in your prompt, it's difficult to specify
which to prioritize. Weights do exactly that by placing more emphasis on one
element over the other.
It works like this:
Place each subject in square parentheses [subject], and add ::1 or ::2 to assign a
weight ratio.
Let's take a girl and a glass of orange juice:
AI Money Code – Emmanuel Cobbi: +237 672754178

Here, the girl is more prominent than the orange juice.

The prompt is “[a girl]::2 and [a glass of orange juice]::1,” and the girl is
twice as important as the orange juice.

Now let's try it the other way around: “[a girl]::1 and [a glass of orange

juice]::2”

Now, the orange juice is clearly the primary subject of the image, while the girl
is pushed to the background. You can play around with the weights to make
either subject more or less prominent.
AI Money Code – Emmanuel Cobbi: +237 672754178

 Stylize Parameter:

The default stylized setting for all images is 100.

You can change this to -- s 0 all the way to -- s 1000. Images with lower
stylization are more accurate to the prompt but can be less creative or aesthetic.
Here's an image with low stylization.

Prompt: “Work desk with plants and books -- s 50”

Here's an image with high stylization.

Prompt: “Work desk with plants and books -- s 1000”
AI Money Code – Emmanuel Cobbi: +237 672754178

More Advanced settings you can try:

These are some of the possible meanings of the terms in your prompt:

 Hyper Detail: This is a term that refers to the level of detail and realism in
a digital image. Hyper detail means that the image has very high
resolution, sharpness, and accuracy, and that it can show fine details and
textures that are not visible in lower quality images.

Hyper detail can also imply that the image is generated by a sophisticated
algorithm or software that can create realistic effects such as lighting,
shadows, reflections, and materials.

 8K: This is a term that refers to the resolution of a digital image or

display. 8K means that the image or display has 7680 pixels horizontally
and 4320 pixels vertically, which is four times the resolution of 4K and
16 times the resolution of Full HD.

 HD: This is a term that refers to the resolution of a digital image or
display. HD stands for High Definition, and can provide better quality
and clarity than Standard Definition (SD), but it is lower than Full HD,
4K, or 8K.
AI Money Code – Emmanuel Cobbi: +237 672754178

 Octane Rendering: This is a term that refers to an engine that can

generate photorealistic images.

 Unreal Engine: This is a term that refers to an engine that can create
interactive 3D content such as games, simulations, animations, etc.

 V-Ray: This is a term that refers to an engine that can generate
photorealistic images using a technique called ray tracing.

 Full HD: This is a term that refers to the resolution of a digital image or
display. Full HD stands for Full High Definition can provide better
quality and clarity than HD, but it is lower than 4K or 8K.

 s5000: This is a term that refers to the size of the image in pixels. s5000
means that the image has 5000 pixels on its longest side. For example, if
the image has an aspect ratio of 4:3 (see example above), then s5000
means that the image has 5000 pixels horizontally and 3750 pixels
vertically.

 uplight: This is a term that refers to the direction of the light source in the
scene. uplight means that the light source is below the objects or camera
angle, creating an effect of illumination from below.

 uplight can create dramatic shadows and highlights on the objects or
characters in the scene.

 q 3: This is a term that refers to the quality of the image compression. q 3
means that the image has a compression quality of 3 out of 10 (where 10
is the highest quality and 1 is the lowest quality).

 stop 80: This is a term that refers to the exposure value of the image. stop
80 means that the image has an exposure value of -80 (where 0 is normal
exposure and negative values are underexposed).

 Exposure value affects how bright or dark the image appears. Higher
exposure value means brighter image and lower exposure value means
darker image.

 w 0.5: This is a term that refers to the weight of the prompt for generating
an AI art. w 0.5 means that the prompt has a weight of 0.5 out of 1 (where
1 is the highest weight and 0 is the lowest weight).
AI Money Code – Emmanuel Cobbi: +237 672754178

 Weight affects how much influence the prompt has on the AI art
generation. Higher weight means more influence and lower weight means
less influence.


I hope this helps you understand how to manipulate and create sophisticated
prompts.

Checkout examples below:

Besides the prompts mentioned above, you could write the prompts with
descriptions for the things you want.

You could put the prompts as follows:

1. A home with windows, doors, porches, awnings, middle of SPACE,

cyberpunk lights, Hyper Detail, 8K, HD, Octane Rendering, Unreal
Engine, V-Ray, full hd - s5000 - uplight - q 3 - stop 80 - w 0.5 - ar 4:3.
2. The planet after human extinction, a new beginning, nature taking back
the planet, harmony, peace, earth balanced - version 3 - s 42000 - uplight
- ar 4:3 - no text, blur, people, humans, pollution.
3. Hyerophant, god light, cinematic look, octane render, under water, -
wallpaper.