Crafting Iterative Prompts for Generating Complex Images with AI: A Beginner's Guide


Artificial Intelligence (AI) is revolutionizing the creative landscape, particularly within the sphere of visual arts. A technique that has opened exciting new avenues of creativity involves using iterative prompts to generate intricate, detailed images. In this tutorial, we're going to explore the craft of formulating effective iterative prompts, guiding your AI model to produce stunning and complex visuals.

It's important to bear in mind that different AI image generators respond differently to the structure of your prompts. For instance, DALLE-2 appears to favor more natural language prompts, as though you're describing the image to a friend or colleague. On the other hand, Midjourney seems to prefer a more methodical and structured approach.

However, there's one common factor - all these AI models pay close attention to the order of words in your prompt. Words at the beginning of your prompt are prioritized over those that follow. Think of it as your prompt starting in high-definition, with the clarity gradually fading as the sentence progresses. The longer your prompt, the more the AI might miss or overlook the latter instructions.

A reliable way to structure your prompt is to divide it into three sections, acknowledging that the AI's attention will taper off as it proceeds. The structure might look something like this: “A large brown dog on a golf course, in the style of an editorial photograph, shaggy and happy, UHD, muted color”. In this example, we start with the main idea (“A large brown dog on a golf course”), followed by the desired style (“in the style of an editorial photograph”), and end with additional personalized details (“shaggy and happy, UHD, muted color”). While this three-part format is not a hard-and-fast rule, it's a good starting point for beginners.

[Glossary Note: Iterative prompts involve providing a series of instructions or 'prompts' to an AI model, with each one refining the output based on the previous results.]

With these considerations in mind, let's jump into the process of crafting an iterative prompt from scratch.

1. Start with a Clear Vision: Just like any artistic endeavor, it's essential to have a clear vision for what you want to convey through your image. Think about the theme, style, composition, or specific elements you'd like to incorporate. This vision will be your north star, guiding you throughout the iterative prompt creation process.

2. Begin with Simple Prompts: In your initial steps, start with a straightforward prompt that lays the foundation for your image. For instance, if you're aiming for an image of a forest landscape, a foundational prompt could be, generate/imagine/create: "A forest landscape."

Note: The AI will be producing four images in a 2X2 grid based upon the prompt(s) I provide. Some AI can generate more options with one prompt, while others fewer. And how those images are displayed for you may vary as well.

AI generated image(s) of a forest landscape.

3. Refine and Add Details: After generating initial versions of your image, observe each output and identify areas needing refinement or additional details. Use your observations to iteratively adjust your prompts, specifying desired changes or enhancements. For instance, you might want to include specific lighting conditions, flora/fauna elements, or atmospheric effects. So, our initial forest landscape prompt might evolve into, “A misty forest landscape with a carpet of ferns in the early morning light.”

AI generated image(s) of a forest landscape featuring ferns and directional lighting.

4. Experiment with Different Styles: When working with AI models, it's beneficial to explore various artistic styles. By incorporating different styles into your prompt variations, you can observe how they impact color palettes, compositions, tone or mood, and overall aesthetics. In this case, since I desire a landscape with a nostalgic yet cinematic quality, my next iterative prompt could be something like: “A misty forest landscape with a carpet of ferns in the early morning light, in the style of epic cinematography, enchanting fireflies, 1970’s color palette.”

AI generated image(s) of a forest landscape featuring cinematic directional lighting.

5. Incorporate Feedback Loops: Feedback loops play a crucial role when working iteratively with AI models. It involves critically reviewing each generated output and adjusting subsequent prompts based on what works well and what needs improvement. [Glossary Note: In this context, a feedback loop refers to the process of using the AI model's output to inform the next input, creating a cycle of continuous improvement.] In this instance, the forest landscape didn't capture the desired 1970’s tone and color palette effectively enough while also lacking prominence for the fireflies.

To improve this, I will rephrase the prompt to emphasize key words/elements that need to be stronger in the image—such as specific details about color and tone—as well as ensuring that the presence of fireflies is more pronounced. Additionally, I will reconsider my choice of words; although terms like "epic" and "enchanting" are creative descriptors, they may be too subjective or vague for achieving precise results. So, I will be adjusting the language around those elements, as well.

To further guide the AI in creating the desired look and feel of the image, I will provide additional context by referencing a specific type of film or photograph that captures the intended aesthetic. With these considerations in mind, an improved version of my prompt might read as follows: "Fireflies shine bright in a misty forest landscape with a carpet of ferns, in the style of dim morning light, 1970’s Polaroid, faded color palette."

Stylized AI generated image(s) of a forest landscape with fireflys hovering in morning light.

6. Break Down Complex Elements: If your image includes several complex elements within the composition, consider tackling each element individually until you feel you’ve got a handle on each. You can then begin to fold those elements into new prompts, combining them in subsequent iterations. This technique allows for more granular control over the final composition.

7. Iterate and Refine Further: The iterative process is all about refining and enhancing your image gradually. Don't shy away from iterating multiple times, experimenting with different prompts, styles, and details until you achieve your vision. Each iteration brings you closer to realizing your creative goal.

8. Embrace Serendipity: While having a clear vision is crucial, don't be afraid to embrace serendipitous moments during the iterative process. Sometimes, unexpected outputs can lead to exciting discoveries or inspire new creative directions that you hadn't initially considered. Such happy accidents can add a unique touch to your final image.

Conclusion: The art of crafting complex images through iterative prompts is a fascinating and rewarding journey. It might seem a bit challenging at first, but remember - every iteration brings you one step closer to your goal. Don't forget to have fun and enjoy the process!

Common Challenges and Tips:

  • Finding the Right Balance: Crafting the perfect prompt often involves balancing specificity and creativity. Too specific, and you might limit the AI's ability to generate interesting outputs. Too vague, and the results might stray far from your vision. Tip: Start with broad prompts, then iteratively add details as needed.
  • Managing Expectations: AI is powerful, but it's not a magic wand. Sometimes, it might not capture your vision perfectly, or it may require many iterations to get close. Tip: Embrace the process and see these moments as opportunities to learn and experiment.