Sponsored by
How to prompt AI image generators for true photorealism
Recently we’ve been having a lot of fun playing with AI image generators that promise to help creative professionals and marketers produce high-quality stock imagery that matches your exact needs or creative brief.
Surprisingly, high-quality imagery can be produced without even needing to sign up for a paid account, since companies like Adobe are offering a limited amount of free AI photo generation to get you started.
For this how-to, we’ve used the free option on Adobe Firefly to look at how images can be generated with minimum fuss and maximum results.
Getting started
As mentioned, we’re using Adobe Firefly for this guide—and you don’t even need to sign up for a paid plan. You will need an account though, so head to firefly.adobe.com and then follow the steps to either sign in or set up a new account.
Once you’re in, you’ll be presented with a prompt box and the instruction to “describe what you want to generate”, as well as a range of image options to help refine your results. Let’s dive in.
Setting your image parameters
Adobe Firefly offers several options that let you control how an image is generated and the format of the final result. This includes the ability to specify which generative model you’d like to use.
The free plan is limited in which models are available, but you can always access Adobe’s commercially safe Firefly Image 3 and Firefly Image 4 models. Signing up to a paid plan will give you more options, with free trials available so you can test before committing.
The other parameters you get to set will depend on which model you choose, but typically include key details such as the aspect ratio. Using Adobe Firefly Image 4, we also found options to set our preferred lighting options, color and tone, and camera angle.
Creating a great prompt
Even the best AI image generator is only as good as the instruction that you give it. A prompt needs to be clear, specific, and worded in a way that the AI model will find easy to interpret. In case you’re not quite sure where to start, Adobe has a range of suggestions to help you craft a thoughtful prompt:
- Simplicity is key: start with a basic phrase that captures the key elements you’d like to see included
- Add plenty of descriptive words to help make the image specific to your needs
- Start small and then add additional detail. For instance, the initial prompt might be for a “Light pink metal water bottle with contrast lid lying on the grass.” From there, you can build in additional details, such as changing the texture of the bottle, adding in weather conditions or prompting the generator to make the image photorealistic.
To create this image, we used the Firefly Image 4 model. One of the things that’s particularly great when using this model is the ability to get help with your prompt. That help comes in two forms: prompt suggestions and prompt enhancement.
Turn on the prompt suggestions toggle, and you’ll see simple suggestions for improving your prompt presented in real time, as you type. In the example below, you can see how it suggested additional descriptive language and context to help make our water bottle image more specific.
Prompt enhancement runs in the background, taking the original prompt that you write up and adding in additional information after you select Generate. In our example it converted the prompt “Light pink matte effect metal water bottle with contrasting purple lid lying flat on the grass, photorealistic, no people in image; studio shot; horizontal. Weather is bright,” to a rather more detailed “A light pink matte effect metal water bottle with a contrasting purple lid lies flat on a lush, emerald green grassy field. The bottle is placed in a studio setting with bright, natural light creating soft shadows. The background is a blurred, vibrant green, enhancing the photorealistic quality of the image. The overall composition is balanced and detailed, highlighting the bottle's sleek design and the natural texture of the grass. The bright, sunny weather adds a cheerful and inviting atmosphere to the scene.”
Having the ability to have AI assist you in improving your prompt is a helpful feature in itself, but it also gives us some clues about what makes an effective set of instructions for any AI photo generator you might be using. Looking at the second prompt, you can see lots of descriptive language, defining everything from the color of the grass to the desired mood and the composition of the image itself.
Using reference images
Another of Adobe’s tips for getting better results with your AI picture generation is to use reference images. Depending on which generative model you’re using, existing images can be used to set the style and composition of the image, or as direct references that show details you want to include—anything from a specific product design to a particular human model that you’d like to use.
This gives far more creative control over the output, and also lets you maintain consistency when generating a set of multiple images. To get the best results using reference images, we found that it was necessary to sign up for a free trial of the Adobe Firefly Pro plan.
This gave us access to the Gemini 3.1 (Nano Banana 2) model, which allows up to six reference images. In our test, it was able to recreate the likeness of our pet rabbit in multiple different scenarios.
Achieving true photorealism
There will be some who say you can always tell the difference between an AI generated image and a genuine photograph. However, assuming the objective is to create high-quality and realistic photographic style images, suitable for use in creative campaigns, image generators certainly make it possible. The key is to have a clear idea of what you’d like your image to include, describe it well and include references for detail.