To the uninitiated, the options within NAI image generators might seem overwhelming, but in essence, there are just two main approaches: text prompting and img2img. This guide will cover both of those approaches in depth, but if you’re new to NAI you may not want to get caught up on all of that just yet-instead, let’s just know that if you’re using NAI, you’ll most likely be working by writing a text prompt or inputting an image file into NAI.
What are NAI Image Generators?
NAI image generators are AI tools that take user prompts to create or modify images. Basically, you write an instruction into an image generator, and it produces an image based on those instructions. These tools are primarily used for character concept art, illustrations, concept art, social media assets, mood boards, etc., though anyone can use it to try out different ideas and concepts to create different styles or rough illustrations.
Why Text Prompts and Existing Photos Matter
The difference between using a text prompt vs. an existing photo is basically the difference between asking an AI to imagine something vs. asking an AI to edit something. A text prompt is just a written description of a photo, for example: “A beautiful forest at sunset.” The AI then does its best to generate the image described in the text.
An existing photo is used as a sort of starting point. You can guide the tool to edit it in the direction that you prefer, for instance, removing the trees and keeping the sunset or vice versa. The starting point makes a huge difference in the control you have over the final outcome as well as how unpredictable it can be.
How a Text Prompt Works
The basis of text prompt generation is a written description of an image. A prompt includes all the necessary details like subject, setting, tone/mood, style, and any important information. A basic prompt looks more like this: “Girl in the garden.” A strong prompt looks more like this: “A young girl standing in the middle of a peaceful garden filled with flowers. Soft morning light. Muted colors. Storybook style.”
Newbies often don’t think prompts need to be that complicated, but the simple ones usually result in plain, boring images. Don’t be afraid to be detailed. Keep it simple if that’s what works for you, but add details until you see them reflected in the image. When you’re writing a text prompt for an AI image generator, consider it to be a short brief vs. a magic spell.
How existing photo image generation works
First, you provide the AI with the photo you want to start from. It analyzes the overall structure of the image (pose, layout, shapes, composition), and generates a new photo, based on a new instruction. This is sometimes referred to as image-to-image. You might use a rough doodle, a plain selfie, a photo of the product, or a photograph of a landscape, to find inspiration from.
AI image generator from existing photo can be used to find variations in style, mood, background, or a more refined image of a rough image you already have.
Text Prompts vs Existing Photos: Main differences
The main difference is where the process begins. With a text prompt you give the AI more liberty, as there is nothing but words for it to start with. The end result might be surprising and interesting, but there is also a larger chance that it won’t look like you expected. You may have to give the prompt a few times before you get something that fits your mental image.
When working with an existing photo, there is more structure in place for the AI. This may keep the result closer to the original pose, shape, or composition. This might be useful when you care about the composition and/or if it is important for the image to still look similar to the original photo. Still, the AI may decide to make different changes, so it won’t be exactly as with a photo editor.
When Should You Write a Prompt?
Use prompts for ideas you have in your mind: when you want to come up with a unique character, fantasy illustration, surreal backdrop, concept for a poster, or a concept collage. Or whenever you don’t have an image or outline to begin with.
In the course of your everyday life, prompts can be helpful for a simple project. Use a prompt to brainstorm a birthday card idea, the theme for a PowerPoint presentation, the look and feel for a room or the character for a story.
Begin with what is the most important; who or what the subject is, what the character is doing, the mood of the scene, etc. Add details later.
When Should You Use a Photo as Input?
Existing photos are more useful if you have an idea of what you want. For example, you might like a pose in a sketch but would rather have it rendered as, say, a watercolor painting, or you might be looking at an image of your room and want to see if new furniture would fit. You might have an existing product photo and want to see how different background options might work for your site.
Using an existing photo also gives you a chance to learn. If you make small changes to an existing photo in response to a prompt, you can see how that difference changes the output. If you use an AI image-to-image generator, for example, try to only change one thing at a time-lighting, background, or style-and you’ll learn better what the AI is responding to.
Things to note for beginners
It’s very powerful, but these tools are not perfect; they may misinterpret your request, include random details, alter facial features, morph hands, or omit part of your input, so don’t consider a single-generation image as the final result.
Pay attention to privacy and consent; it is wise to avoid uploading images of yourself and others, and try to use photos of faces, body parts, and personal content without the consent of the people or images themselves; additionally, consider copyright issues if you intend to publish or distribute the final result, as AI tools should not be employed to plagiarize someone else’s work and impersonate other persons’ images.
AI tools are only meant to be a helper to your creativity, not to replace it.
Tips for Improved Outcomes
Define your objective first. Before generating images, determine what you want the images to be used for. For instance, do you want them to serve as inspiration, a part of a personal project, a starting idea, or something for public display? Knowing this will tell you how comprehensive your prompt needs to be and whether your prompt should consist of text or a reference photograph.
Simplify your initial attempts. Begin your prompts with fewer words, see what comes up, and then enhance it. Save some results that are useful even if not flawless, so that you have a reference point in mind. You might find you want to take an image that is pretty close and add or change just one aspect of the prompt, rather than completely reworking it.
What Is Next for NAI?
NAI may well become even more user-friendly and accessible down the road. The models will probably get better at interpreting natural language commands, following visual prompts and blending text input with already existing photos. It’s likely going to take fewer adjustments to get good results from NAI when you are learning.
On the other hand, there will likely be greater emphasis on using NAI responsibly. As NAI becomes easier to use, users will be better served by thinking about whether they have the other person’s permission, whether they can be truthful with what they show or whether they are providing any context for it. A realistic AI picture can be a fun tool or helpful, but it can also be confusing when it’s used with no explanation.
Final thoughts
You can interact with the NAI image generators with text prompts or with existing photos. The former is most useful when you’re trying to generate something out of your imagination, while the latter is most useful when you want to influence the AI with a visual cue.
The best way to approach AI image generators as a beginner is to try them out gradually. Begin with a specific idea in mind, use short, simple text prompts, avoid using personal images as starting points, and carefully review the results. AI image generators can be a useful tool for creating new and original ideas, planning projects, and getting inspiration, so long as you understand what they do and don’t do.
