Adobe Firefly

basicwidespread

Adobe Firefly is Adobe's family of artificial intelligence tools that generate images, video, audio, and graphics, integrated into Photoshop and other company software.

In plain terms

Firefly is like a very fast drawing assistant at a graphic designer's desk: you describe in words what you want to see, and within a few seconds it presents you with several variations. You can create a new image from nothing, but it is also useful for removing something from an existing photo, expanding the background, or replacing a detail.

Most people encounter it not as a separate program but within familiar Adobe tools: behind Photoshop's "generative fill" feature, Illustrator's vector graphics creator, or Adobe Express templates. There is also a standalone web and mobile interface where you can create images, videos, and audio.

What distinguishes it most from other image generators is that Adobe trained its own models on images licensed from Adobe Stock, open-license content, and content whose protection period has expired. Therefore, the company promises that the results can be used safely for business purposes, meaning there is less copyright risk than in systems trained on images collected indiscriminately from the internet.

How it works

You write the request (prompt) in everyday language, for example: "autumn forest in morning mist, watercolor style". You can also provide a reference image for the style or composition.
Like most modern image generators, Firefly's image models operate on the diffusion principle: they start from random 'image noise' and refine it step by step until an image matching your description emerges.
If you edit an existing image, you select the area to be modified, and the system regenerates only that part so that the lights, shadows, and perspective fit the surroundings.
You receive multiple versions to choose from, which you can then further refine using standard editing tools. Generation consumes so-called generative credits depending on your subscription. Some packages offer unlimited basic image functions, while premium features (such as video) require credits.
The system writes content authenticity data (Content Credentials) into the finished file, indicating that the image was created with AI.

An everyday example

You photograph your family at Lake Balaton, but a stranger and a trash can stand in the background. In Photoshop, you draw around them, click generative fill, and Firefly 'paints' water and reeds matching the shore in their place.

Who is affected

marketerYou can create campaign images, banner variations, and social media content in minutes instead of hours. If you use Adobe's own models, legal risk is lower. → Marketing

employeeIf you work as a graphic designer, photographer, or videographer, it is integrated into your daily tools and reshapes which tasks remain manual. → Work & HR

entrepreneur, executiveEven as a small company, you can produce acceptable-quality visual material, but you must know what the license covers and what it does not.

parent, student, teacherEasily accessible for school posters, presentations, or creative tasks, and a good opportunity to discuss how to recognize AI-generated images. → Education & Family

A common misconception

MythWhat Firefly generates is legally completely risk-free, and the copyright of the image automatically becomes yours.

In realityAdobe did train its own models on licensed material, and it provides legal protection for certain corporate packages, but this does not apply to every subscription or every model. Whether a purely AI-generated image receives copyright protection is still debated in many countries.

Limits and risks

Generated images may contain errors: distorted hands, nonsensical text, strange details. Always review the result before using it.
Usage is partly credit-based: the quota can run out, and video generation and premium features cost far more credits than a simple image.
On the Firefly interface, models from other companies can also be selected, and these do not carry the same training data and legal security promises as Adobe's own models.
The content authenticity label can be removed or lost if the image is shared as a screenshot or through certain platforms, so it does not by itself prove the origin of an image.

Similar, but not the same

Photoshop – Photoshop is the image editing program, while Firefly is the generative AI model family that powers some of its features (such as generative fill).
Midjourney – Midjourney is a standalone image generator focused mainly on artistic visuals, while Firefly is integrated into editing software and emphasizes commercial usability.
Adobe Sensei – Sensei is Adobe's previous, mainly analytical and automating AI technology (for example, smart selection), while Firefly creates new content.

Related concepts

Broader

generative AI, image generator, generative media

Narrower

generative fill, generative credit

On the edge

diffusion model, Midjourney, Stable Diffusion, content authentication, prompt

A brief history

Adobe introduced the beta version of Firefly in March 2023. Later that year, first as a beta, it integrated generative fill into Photoshop, then opened it for commercial use in September. It announced its video-generating model in autumn 2024, and it became publicly available in early 2025. In 2025, the standalone Firefly application also incorporated models from partner companies.

Where things stand

As of late 2026, Firefly is a built-in part of Adobe's creative applications and generates images, vector graphics, video and audio. In the standalone Firefly app, models from partner companies are available alongside Adobe's own models.

Status: 2026 November

Also known as

Adobe Firefly · Firefly · Firefly AI · Adobe Firefly MI · Firefly Image Model · Firefly Video Model · Adobe generative AI · Generative Fill

Sources

Wikipedia – Adobe Firefly

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.