This content originally appeared on HackerNoon and was authored by Obinna Mbajunwa
You’ve probably heard countless times that when it comes to prompting—whether on ChatGPT, MidJourney, Synthesia, Runway ML, or any other AI tool—being as precise as possible is key. The first step toward precision is detail. AI, as we know it, is not a mind reader; it works solely based on the information you provide. Only with clear, detailed input can you achieve the best results.
\ This principle also applies when generating precise images. However, in this case, when prompting, you need to think like a photographer, a designer, and a writer all at once.
\ In this article, we’ll explore how to write effective prompts for Google's AI Test Kitchen ImageFX to get accurate results.
\
First of all, the Don't of AI Test Kitchen Image FX
- ImageFX does not allow queries that could lead to outputs involving minors or certain prominent individuals.
- It also discourages queries that result in images portraying abuse, violence, self-harm, or any content promoting hatred.
\
Why Does Precision Matter?
Precision can mean the difference between a flawed result and a perfectly executed output, especially when generating images. Text prompts might offer vague summaries, but poorly structured image prompts can drastically affect the outcome. If, like me, you’re creating for a non-Caucasian audience, you must be extremely precise. A vague prompt will likely produce generic, unusable results. However, when done correctly, ImageFX excels at generating region-specific outputs tailored to your needs.
Crafting Effective Prompts: A Step-by-Step Guide
When crafting an image prompt, keep these three key components in mind:
1. Who Is Your Subject Matter?
Suppose you’re conceptualizing an image of a "woman in a coffee shop." In this case, the woman is the subject. It’s essential to clearly define your subject, depending on the audience and your intention.
\ See the difference between both prompts.
\ Prompt 1: A woman in a coffee shop.
\
\ Prompt 2: A plump Nigerian woman sitting in a coffee shop.
\
\ \n 2. Be Descriptive and Detailed
\ In other words "show workings." What actions are the characters performing? What time of day is it? What is the mood? Describe the setting, the characters’ outfits, any secondary subjects, and the overall atmosphere.
\ Example:
A plump Nigerian woman sitting in a coffee shop, sipping coffee. On the table is a book and an iPhone and seated beside her is a Pomeranian.
\
\ 3. Be Specific About the Composition and Style
This is where you need to think like a designer. Consider the layout and style. Do you want an animated style? If so, what kind? Do you want it to be animated and if yes what kind of animation - Traditional Animation (2D, Hand-Drawn Animation) or 3D Animation (CGI)? Luckily, if you are looking to adopt a particular style of animation that you like, ImageFX allows you to experiment with different animation styles.
\ Prompt
A plump Nigerian woman sitting in a coffee shop, sipping coffee. On the table is a book and an iPhone and seated beside her is a Pomeranian. They are all in Pixar animated style.
\n
\ A plump Nigerian woman sitting in a coffee shop, sipping coffee. On the table is a book and an iPhone and seated beside her is a Pomeranian. They are all in Simpson's animated style.
\
\ To get into a photographer's mindset, you have to consider the angles, the lighting, and the focus of the shot.
Prompts with Various Angles:
1. Close-up (Focus on the Woman and Coffee Sip)
Close-up shot of a plump Nigerian woman lifting her coffee cup toward her lips, taking a slow sip. Her face shows a relaxed expression. On the table in front of her, a book lies open next to an iPhone, partially in view. A Pomeranian sits quietly beside her, peeking into the frame.
\
\
2. Over-the-shoulder (Focus on the Table and Scene)
Over-the-shoulder shot from behind a plump Nigerian woman, capturing the scene from her perspective. The table shows an open book and her iPhone lying beside it. The Pomeranian is seated next to her, with the coffee shop atmosphere visible in the background.
\
\
3. Wide-angle (Full Room View of the Scene)
Wide-angle shot showing the entire coffee shop scene: a plump Nigerian woman sitting at a table, sipping coffee, with her book and iPhone lying in front of her. A Pomeranian is seated by her side, looking calm. The café setting includes soft lighting, other customers, and a warm ambiance.
\
\
4. Bird’s-eye View (Top-down Perspective of the Table)
Bird’s-eye view from directly above the table shows the entire scene: an open book, an iPhone, and a coffee cup on a table. A Pomeranian is seated beside a plump Nigerian woman, while she is sipping her coffee, creating a peaceful, content moment captured from above.
\
\
5. Medium Shot (Balanced View of the Woman, Table, and Pomeranian)
Medium shot from the side, showing a plump Nigerian woman comfortably sitting in the coffee shop, sipping her coffee. A table is visible with the open book and iPhone laid out, and a Pomeranian is seated next to her, calmly watching the surroundings. The warm lighting of the coffee shop adds to the ambiance. \n
\
6. Side Profile (Highlighting the Woman and the Pet)
Side profile shot of a plump Nigerian woman sitting at a table, sipping her coffee, with a book open in front of her and the iPhone placed nearby. A Pomeranian is seated beside her, either looking toward her or gazing ahead. The coffee shop's interior provides a soft, cozy backdrop. \n
\
7. Close-up (Focus on the Pomeranian)
Close-up shot of a Pomeranian seated on the chair beside a woman, looking calm and content. In the blurred background, a plump Nigerian woman can be seen sipping her coffee with the book and iPhone on the table.
\
\
9. Overhead Angle (Showcasing Coffee, Book, and Phone)
Overhead angle shot that captures a plump Nigerian woman’s hand holding the coffee cup as she lifts it to take a sip. A book lies open on the table with an iPhone next to it, creating a top-down view of the table setup. A Pomeranian is just visible seated beside her, rounding out the scene.
\
Tips for Getting the Best Results on ImageFX
- Experiment with Keywords: Don’t hesitate to try different keyword combinations. ImageFX might be restrictive, but with the right prompts, you can achieve close-to-ideal results.
- Maintain Consistency: When creating a series of images, keep the angle and style consistent. Only change the subject when necessary.
- Use Identifiers: When generating multiple images with the same subject, give the subject unique identifiers to maintain flow.
Troubleshooting Common Issues with ImageFX
Query: "I’m a Nigerian creating for a Nigerian audience, but some details are off." Response: AI isn’t a mind reader. You need to adjust your descriptors for more accurate results.
\ Example:
Prompt 1: A Nigerian conductor in front of a bus.
\ Prompt 2: A Nigerian Lagos conductor wearing shorts and a tank top in front of a yellow Danfo bus (Volkswagen T3).
\
Conclusion
While you don’t need to be a prolific writer to craft effective prompts, you do need to take the time to clearly and descriptively outline what you want. Don’t be discouraged if the result isn’t perfect on the first try—keep experimenting, adding or removing keywords as needed.
\ And with that I leave you with these images I conceptualised using ImageFX. Enjoy.
\
This content originally appeared on HackerNoon and was authored by Obinna Mbajunwa
Obinna Mbajunwa | Sciencx (2024-10-25T19:00:15+00:00) How to Write Precise Prompts on Google’s AI Test Kitchen. Retrieved from https://www.scien.cx/2024/10/25/how-to-write-precise-prompts-on-googles-ai-test-kitchen/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.