Safe for Beginners! How to Write Prompts That "Get Through" to Image Generation AI
Image generation AI—does it seem a bitdifficult? Are you holding yourself back because you assume that, and finding it hard to get started?
It might feel difficult for those who have never used image generation AI or are beginners, but that is not the case at all.
Image generation AI creates wonderful pictures in an instant by having you convey the image of "I want to draw this!" through words (called prompts).
However, what kind of words (prompts) and how should you convey them so that the AI understands your image?
The key is to give the AI concrete and easy-to-understand instructions to the AI.
In this note, I will introduce a simple method of thinking about prompts by dividing them into four parts, without using difficult technical terms as much as possible.
Even if you are a complete beginner to generative AI, I have prepared several examples so that you can try them out immediately after finishing this note.
Let's grasp the tips for writing prompts together and challenge yourself to create your own original images!
As of May 5, 2025, you can use ImageFX provided by Google for free, so please use it for practice. It supports Japanese prompts, so you can create images by inputting Japanese.
*A Google account is required for use.
1. Main Subject Part | Let's Decide the "Face" of the Picture!
The first thing to do is to decide on the thing you want to stand out the most in the picture, in other words, the " main subject ".
Clearly tell the AI who or what is the center.
■ Example:
「white owl」「astronaut」「old castle」「smiling girl」「sparkling jewel」
2. Location/Time Part | Let's Set the Stage!
Next, tell the AI "where" the main subject is and "when" the scene takes place.
By adding the background and time of day, the world of the picture will expand significantly.
■ Examples of locations:
「lakeside at sunrise」「futuristic city with neon lights」「inside a quiet library」「under a row of cherry blossom trees in full bloom」
■ Examples of times:
「midsummer afternoon」「dusk after the rain」「starry night」「foggy early morning」
■ Examples of combinations:
「summer night, town with fireworks」「seaside where the sun sets」
3. Atmosphere Parts | Let's Create a Mood!
These are parts that determine the overall "atmosphere" of the picture. You tell the AI what color scheme, brightness, and general feel (vibe) you want the picture to have. You can also convey your preferences for the art style here.
■ Examples of colors and brightness:
"soft pastel colors", "sparkling light", "calm monochrome", "vivid colors", "warm light like a fireplace"
■ Examples of emotions and textures:
"cheerful atmosphere", "a slightly lonely feeling", "fluffy texture", "smooth surface"
■ Examples of art styles:
"anime style", "watercolor touch", "photorealistic", "oil painting style", "fantasy art"
4. Other Parts (Add as you like) | Add Even More Detail!
These parts are not mandatory, but you can add them when you want to get closer to your ideal image. You can tell the AI about the image shape (aspect ratio), things you don't want drawn (negative prompts), and other specific requests.
■ Image shape (aspect ratio):
"landscape image (16:9)", "square image (1:1)", "portrait image for smartphone wallpaper (9:16)"
■ Things you don't want drawn (negative prompt):
"no text", "no people"
As a practical example, let's create an image with the following prompts.
Prompt 1: "Inside a quiet library"
Prompt 2: "Inside a quiet library, do not draw a desk"
Prompt 1 Example: "Inside a quiet library"
Prompt 2 Example: "Inside a quiet library, do not draw desks"
By giving specific instructions like "do not ~" or "exclude ~," it becomes easier for the AI to understand.
■ Others (Composition/Visual Effects):
"Wide-angle composition", "Close-up of face", "Simple background"
Examples of combining prompts | Good prompts
Just by combining the parts we've covered so far in order, you can create a prompt that is easy for the AI to understand!
■ Example 1:
"White owl, by a lake at sunrise, soft pastel colors, landscape image (16:9)"
Subject: White owl
Location/Time: By a lake at sunrise
Atmosphere: Soft pastel colors
Others: Landscape image (16:9)
*In ImageFX, you can specify aspect ratios like landscape or portrait in the settings, so there is no need to include "landscape image (16:9)" in the prompt. You only need to include instructions for "landscape image (16:9)" when using an AI image generator that does not have aspect ratio settings, such as DALL-E 3 integrated into ChatGPT.
■ Example 2:
"Smiling girl, under a row of cherry blossom trees in full bloom, midsummer afternoon, anime style, sparkling light, square image (1:1)"
Subject: Smiling girl
Location/Time: Under a row of cherry blossom trees in full bloom, midsummer afternoon
Atmosphere: Anime style, sparkling light
Others: Square image (1:1)
■ Example 3:
"A neon-lit futuristic city, a starry night, photorealistic, slightly lonely atmosphere, vertical image (9:16), no people"
Subject: (None. The scenery is the subject)
Location/Time: Neon-lit futuristic city, starry night
Atmosphere: Photorealistic, slightly lonely feeling
Other: Vertical image (9:16), do not include people
By being conscious of these parts and combining them, you can create specific and easy-to-understand instructions. If you use the same prompt, the AI is much more likely to recreate an image with a very similar atmosphere.
Example of a bad prompt
Vague specifications or contradictory instructions will confuse the AI.
■ Example: "cute beautiful cat and dog and rain and sunny"
What is the subject?: A cat? A dog? Both? The AI will be confused.
What is the weather?: Rain and sun rarely happen at the same time. These are contradictory instructions.
What about just "cute" or "beautiful"?
Because it is not specific about how it is cute or beautiful, the AI struggles to interpret it, and the resulting images tend to have a different atmosphere every time.
Let's Try It! Practice Corner
Now, let's use what you've learned so far to actually write a prompt!
Try it out while taking notes!
Once your prompt is complete, let's try generating an image using a tool like ImageFX.
Decide on the main subject: What do you want to be the center of the image?
Choose one animal, character, object, etc. (e.g., penguin).Decide on the location and time: Where is your subject? What time of day is it? Let your imagination run wild. (e.g., a city street at night covered in snow)
Decide on the atmosphere: What kind of colors, lighting, or art style do you want to see? (e.g., quiet and calm atmosphere, watercolor style)
(If necessary) Add other details: Are there any image shapes or specific things you don't want drawn? (e.g., square image, do not draw signs)
Put them together to complete the prompt!: Try connecting the words you thought of in steps 1-4 into a single line, separated by commas (,) or spaces.
Example: "penguin, city street at night covered in snow, quiet and calm atmosphere, watercolor style, square image, do not draw signs"Try entering it into the AI: Try entering the prompt you created into an actual AI image generation tool (there are many free ones available). It's exciting to see what kind of image comes out!
Compare and adjust: If the resulting image is different from what you imagined, try changing just one word in the prompt, or adding or removing elements, and try again.
For example, change "watercolor style" to "anime style," or "quiet and calm atmosphere" to "fun and lively atmosphere."
Conclusion
How was it? Were you able to create an image just as you imagined using AI?
Did you feel excited while writing the prompt, wondering what kind of image it would become?
At first, you might have moments where you struggle because you can't quite create the image you imagined, but that's okay.
By repeatedly being conscious of the four parts and trying out different words and combinations, you will surely get the hang of it.
Please go ahead and create many wonderful images that are uniquely yours.
