How to write prompts like a pro
In the modern world, creativity has taken a new hit and using AI technology we can take our Art to new horizons.
Zoomerang is known for its remarkable AI features which will lead your work to astonishing heights.
While many of us turn to AI to get the desired outcomes, it's crucial to recognize that these outcomes are closely tied to the quality of our commands.
Therefore, if we know how to write accurate prompts, we will get better and precise results.
In this article, we will dive into the techniques and effective methods of communicating with AI tools.
Hence, let's break this into specific keyword areas:
Subject
Medium
Style
Resolution
Color
Lighting
Extra details
Limiting the variation
Technique/syntax
The Subject means that you need to specify what you see in the generated media (image/video)
Suppose we aim to generate an image of “Superhero” a beginner will just write Superhero in the prompt area.
However, this gives a huge range to the imagination. How do you wish the Superhero to look? What background you want to see? Which superpower does he/she have? From which movie/comics is superhero? Is he/she standing, running, sitting?
The system can't read our minds, that is why we need more specification.
A common trick for human subjects is to use celebrity names. They have a strong effect and are an excellent way to control the subject’s appearance.
Let's try this example:
Note: the feature used for the demo is Photo AI
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look
The Medium refers to the material used to make the AI result. Here you can specify some examples of artwork, like illustration, 3d, photography etc.
We will add digital painting to our prompt:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting
We can see that the photo is changed to a digital painting. Let's continue.
The style (apart from the predefined Styles by Zoomerang) refers to the artistic style of the image. For instance, impressionist, surrealist, renaissance, expressionist, pop art, sci-fi etc.
Let's now add this style to our demo:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi
The resolution works wonders when you aim to determine the sharpness you desire in the final result. For example, highly detailed, sharp focus etc.
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed
We can see significant changes in the demo, let's hop in to the next category.
Here, you can change the main colors of the desired result by using color related words, those will shape the dominant color within the objects of your generated file.
You can add more complex words to the prompt like iridescent silver, radiant gold, beaming color, etc.
Assume, we wish to add radiant gold to our demo, so the prompt will look like this:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold
By taking a break, please, keep in mind that Zoomerang offers a variety of predefined styles that were made for you.
This post is for understanding major points on writing effective prompts. Also, you don't need to use all the categories every time, you are free to work with them as your preferences!
If you ask any photographer the key of a good image, they will tell you first and foremost is the lighting. So, we can have different prompts here, like: soft lighting, natural light, moonlight, cinematic, etc.
Suppose we wish our Superhero to have a cinematic lighting:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold, cinematic lighting
Extra details serve as enhancements that can transform the generated file. These prompts can have texts like, stunningly beautiful, dreamy, romantic etc. also, they can be region specification keywords. For this instance, we can add a sword in the right hand.
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold, cinematic lighting, stunningly beautiful, a sword in the right hand
Now, we can confidently say that we've addressed all the previously mentioned areas, and our prompts couldn't be more ideal!
However, remember there's always room for crafting more precise prompts.
As a machine, AI can have its own interpretation of certain objects. So, for example, when we say only one keyword, the AI will get us numerous unrelated results.
This underscores the importance of precision, when writing prompts!
Let's explore what happens when we start with a single prompt and then add more specific details.
The other way to get better results is to adjust the keyword weight, which is done by a simple syntax.
Here is an example of a syntax which increases the weight of the written keyword (prompt)
(keyword): 1.1
((keyword)): 1.21
(((keyword))): 1.33
Here, the greatest weight is (((keyword))): 1.33
The more () you write, the more it indicates the importance of your keyword as well as the value, (number) again increases the importance of it.
Note: no need to add more than three Parenthesis if you wish to get better emphasis on the keyword, just adjust the value.
This works the same way when trying to decrease the keyword weight. So, if you wish to decrease the weight, just change the value and add ().
The more Parenthesis you write, the more it will decrease. This is followed by the value, and it should be less than 1.0.
Here is an example of a syntax which decreases the weight of the written keyword (prompt)
(keyword): 0.9
((keyword)): 0.81
(((keyword))): 0.73
Here, the least weight is (((keyword))): 0.73
To visualize all of this, let's write a prompt and see what happens when we use both of those syntaxes.
Last tip, please be aware, that stereotyping and bias is a big topic in AI models so when writing prompts make sure that if you want to avoid stereotypes, you need to draft your prompt as accurate as you can!
We hope that we could cover most question connected with prompts and that you enjoyed the reading!
Zoomerang is known for its remarkable AI features which will lead your work to astonishing heights.
While many of us turn to AI to get the desired outcomes, it's crucial to recognize that these outcomes are closely tied to the quality of our commands.
Therefore, if we know how to write accurate prompts, we will get better and precise results.
In this article, we will dive into the techniques and effective methods of communicating with AI tools.
Hence, let's break this into specific keyword areas:
Subject
Medium
Style
Resolution
Color
Lighting
Extra details
Limiting the variation
Technique/syntax
Subject
The Subject means that you need to specify what you see in the generated media (image/video)
Suppose we aim to generate an image of “Superhero” a beginner will just write Superhero in the prompt area.
However, this gives a huge range to the imagination. How do you wish the Superhero to look? What background you want to see? Which superpower does he/she have? From which movie/comics is superhero? Is he/she standing, running, sitting?
The system can't read our minds, that is why we need more specification.
A common trick for human subjects is to use celebrity names. They have a strong effect and are an excellent way to control the subject’s appearance.
Let's try this example:
Note: the feature used for the demo is Photo AI
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look
Medium
The Medium refers to the material used to make the AI result. Here you can specify some examples of artwork, like illustration, 3d, photography etc.
We will add digital painting to our prompt:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting
We can see that the photo is changed to a digital painting. Let's continue.
Style
The style (apart from the predefined Styles by Zoomerang) refers to the artistic style of the image. For instance, impressionist, surrealist, renaissance, expressionist, pop art, sci-fi etc.
Let's now add this style to our demo:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi
Resolution
The resolution works wonders when you aim to determine the sharpness you desire in the final result. For example, highly detailed, sharp focus etc.
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed
We can see significant changes in the demo, let's hop in to the next category.
Color
Here, you can change the main colors of the desired result by using color related words, those will shape the dominant color within the objects of your generated file.
You can add more complex words to the prompt like iridescent silver, radiant gold, beaming color, etc.
Assume, we wish to add radiant gold to our demo, so the prompt will look like this:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold
By taking a break, please, keep in mind that Zoomerang offers a variety of predefined styles that were made for you.
This post is for understanding major points on writing effective prompts. Also, you don't need to use all the categories every time, you are free to work with them as your preferences!
Lighting
If you ask any photographer the key of a good image, they will tell you first and foremost is the lighting. So, we can have different prompts here, like: soft lighting, natural light, moonlight, cinematic, etc.
Suppose we wish our Superhero to have a cinematic lighting:
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold, cinematic lighting
Extra details and region specific prompts
Extra details serve as enhancements that can transform the generated file. These prompts can have texts like, stunningly beautiful, dreamy, romantic etc. also, they can be region specification keywords. For this instance, we can add a sword in the right hand.
superhero, goddess, DC, ability to fly, super strong, detailed outfit, superwoman, Gal Gadot, detailed clothing, real life look, digital painting, sci-fi, sharp focus, highly detailed, radiant gold, cinematic lighting, stunningly beautiful, a sword in the right hand
Now, we can confidently say that we've addressed all the previously mentioned areas, and our prompts couldn't be more ideal!
However, remember there's always room for crafting more precise prompts.
Limiting the variation
As a machine, AI can have its own interpretation of certain objects. So, for example, when we say only one keyword, the AI will get us numerous unrelated results.
This underscores the importance of precision, when writing prompts!
Let's explore what happens when we start with a single prompt and then add more specific details.
Technique/Syntax
The other way to get better results is to adjust the keyword weight, which is done by a simple syntax.
Here is an example of a syntax which increases the weight of the written keyword (prompt)
(keyword): 1.1
((keyword)): 1.21
(((keyword))): 1.33
Here, the greatest weight is (((keyword))): 1.33
The more () you write, the more it indicates the importance of your keyword as well as the value, (number) again increases the importance of it.
Note: no need to add more than three Parenthesis if you wish to get better emphasis on the keyword, just adjust the value.
This works the same way when trying to decrease the keyword weight. So, if you wish to decrease the weight, just change the value and add ().
The more Parenthesis you write, the more it will decrease. This is followed by the value, and it should be less than 1.0.
Here is an example of a syntax which decreases the weight of the written keyword (prompt)
(keyword): 0.9
((keyword)): 0.81
(((keyword))): 0.73
Here, the least weight is (((keyword))): 0.73
To visualize all of this, let's write a prompt and see what happens when we use both of those syntaxes.
Last tip, please be aware, that stereotyping and bias is a big topic in AI models so when writing prompts make sure that if you want to avoid stereotypes, you need to draft your prompt as accurate as you can!
We hope that we could cover most question connected with prompts and that you enjoyed the reading!
Updated on: 27/10/2023
Thank you!