Generate realistic photos with a prompt formula in Firefly

Start with reference images and build out a prompt formula that controls pose, framing, lighting, and texture.

Published April 13, 2026

00:01

Introduction and meet Jesús Ramirez

[Jesús Ramirez]: Learn the formula to generate realistic photos in Adobe Firefly.

My name is Jesús Ramirez.

In this video, you will learn to generate realistic images using a simple prompt formula you can use for any project.

The goal is to take a product shot, add a portrait reference, and generate a realistic product shot with the same consistent person in multiple environments.

To start, go to the Firefly website at firefly.adobe.com.

From the Prompt box, choose Image from the dropdown.

Then click the More button to access the Generate screen.

This gives you access to the Properties panel on the left, which displays all the settings for the currently selected model.

From the Model dropdown, select the latest Gemini model - Gemini 3.1 (w/ Nano Banana 2) in this case - which is Firefly's most capable partner model for generating realistic images.

You can also select an Aspect ratio - Auto is default, which lets Firefly decide the output - but in this case, choose Widescreen (16:9).

Then select your output's Resolution.

Currently, you can choose between 1K and 2K.

I'll choose 2K.

01:05

Start with a product image and a simple prompt

Next, upload your product image.

Click the Image upload icon in the Properties panel.

You will be able to upload images from your Adobe Cloud Storage and Your device.

I'll choose Your device and select this product photo of a sweater.

Internally, Firefly labels each uploaded image in order, so your sweater becomes Image 1.

Keep in mind that you can reference it in any prompt simply by writing Image 1, and Firefly knows exactly what you mean.

For the first generation, we'll keep the prompt simple.

This is the formula we will use for the first example: subject plus location and context.

Using natural language, tell Firefly who it is and where they are.

The prompt on screen is what I will use.

Notice the color coding that matches the formula.

To save time, I'll simply paste the prompt into the Prompt box at the bottom.

Then I'll click the blue Generate button to create my image.

In a few moments, Firefly will give us a realistic photo.

The sweater detail comes entirely from the reference image, and the prompt established who is wearing it and where.

This is a convincing result, but who is that person?

If you just need a single realistic image, this works perfectly.

02:17

Add a portrait for character consistency

But if you're creating a series, uploading a portrait gives you character consistency across every generation.

To do so, I'll click on the Image upload icon again, and from my device, I'll choose a person's photo.

This becomes Image 2 and you can reference her directly in the prompt.

The prompt on screen is what I will use.

Notice that I'm referencing the woman in Image 2.

After I Generate, Firefly will give us a similar scene with the person in the reference photo, and we'll continue using her across every example in this video.

Now, this result is great, but Firefly made a lot of decisions for us.

In a real photo shoot, you direct your subject, frame the shot, and set the lighting.

And that's exactly what we're adding to the prompt formula next to give you more control over your generated images.

We'll use three new elements in this prompt: action, composition, and lighting.

This is the prompt that we'll use next.

She is now in a professional setting.

We want to give the model more direction on what to generate.

Action sets how she's standing.

Composition sets the framing.

And lighting sets how to illuminate the scene.

I'll Generate this prompt, and this is my result.

Notice the pose.

One hand resting on the desk facing the camera, the framing from the hips up, keeping the full sweater visible, and the warm ambient light keeps no harsh shadows.

Next, let's take the prompt one step further.

03:41

Control action, composition, lighting, and texture

For the third generation, we're going for a nightlife editorial look.

This time, we'll add three elements to the formula: style, camera, and material.

This is the prompt we'll use next.

Style is your creative direction.

It covers that color grade, the editorial feel, and the scene's atmosphere.

Camera sets the photographic settings.

If you're a photographer, this is where your skills translate directly.

The Gemini model understands camera settings the way you do: focal length, aperture, depth of field.

Just use natural language to describe what you want.

And texture sets how things should look up close.

It's how you define the physical surface of everything in the frame, like the fabric and the skin.

I'll Generate.

And the results are fantastic.

Look at the depth of field, the rim light on the side, and the texture on her skin and sweater.

Each one came from a single line in the prompt.

Now try using any of these prompt formulas to generate realistic photos.

Start with the simplest version and build from there.

Use as much of the formula as the task requires.

And remember, not every generation will be the result you want on the first try.

Just like with a real photo shoot, the best photos come from adjusting, iterating, and trying different approaches until the image feels right.

Again, my name is Jesús Ramirez.

Thank you for watching.

What you’ll learn

Start with a product image and a simple prompt

Upload a product photo, select a generative model like Gemini 3.1, and write a prompt that pairs a subject with a location. Firefly generates a realistic scene built around your reference image.

Add a portrait for character consistency

Upload a second reference photo of a person to keep the same face across every generation. Reference both images in your prompt to produce a series with a consistent character and product.

Control action, composition, lighting, and texture

Expand your prompt with direction for pose, framing, camera settings, and material detail. Each element translates into visible changes — from depth of field and rim lighting to fabric and skin texture.

Instruction by: Jesús Ramirez

Cookie Settings

General information

Cookies are small text files stored by your web browser when you use websites. There are also other technologies that can be used for similar purposes like HTML5 Local Storage and local shared objects, web beacons, and embedded scripts. These technologies help us do things like remembering you and your preferences when you return to our sites, measure how you use the website, conduct market research, and gather information about the ads you see and interact with.

If enabled:

We can improve your experience by tailoring the site and the content to things we think might be of interest
We can better keep track of your preferences — like what language you prefer to use
We will better understand your likely interests so we can provide you more relevant Adobe ads and content on non-Adobe websites and in non-Adobe apps
It will help us improve the performance of our website and those of our partners who use the Adobe Experience Cloud

If disabled:

We won’t be able to remember you from session to session so the experience may not be tailored to your interests
You’ll still have access to the content of the site but certain features that depend on cookies may not function
You’ll still see ads, they just may not be as relevant to you

General information

Cookies are small text files stored by your web browser when you use websites. There are also other technologies that can be used for similar purposes like HTML5 Local Storage, web beacons, and embedded scripts. These technologies help us do things like remembering you and your preferences when you return to our sites, measure how you use the website, conduct market research, and gather information about the ads you see and interact with.

If enabled:

We can improve your experience by tailoring the site and the content to things we think might be of interest
We can better keep track of your preferences — like what language you prefer to use
We will better understand your likely interests so we can provide you more relevant Adobe ads and content on non-Adobe websites and in non-Adobe apps
It will help us improve the performance of our website and those of our partners who use the Adobe Experience Cloud

If disabled:

We won’t be able to remember you from session to session so the experience may not be tailored to your interests
We’ll still count your use of our site and services
You’ll still have access to the content of the site but certain features that depend on cookies may not function
You’ll still see ads, they just may not be as relevant to you

View transcript

What you’ll learn