Adobe Firefly

Pros and cons

What Adobe Firefly does

Adobe Firefly is an AI image generation and editing system. Unlike [Midjourney](/tools/midjourney) or DALL-E 3, which generate images from scratch, Firefly's most-used capabilities are editing tools: generative fill (removing or replacing objects in a photo), generative expand (extending the canvas of an existing image), and text-to-image generation for creating new visuals from a prompt.

Firefly is integrated into Adobe Photoshop and Illustrator as native features, not a separate tool. In Photoshop, you select an area, type a description, and Firefly fills it in. Remove a person from a background, extend a sky, replace a product label — these tasks work through the same selection-and-prompt interface you already use. For designers and retouchers who live in Photoshop, Firefly removes the need to switch to a separate AI tool for editing tasks.

The defining characteristic of Firefly is what Adobe calls commercial safety. Firefly's models were trained exclusively on Adobe Stock images (licensed content), openly licensed content, and public domain material. This means the outputs are not derived from copyrighted work without permission — a critical requirement for agencies, brands, and enterprise marketing teams that face legal exposure from using AI-generated images with unclear IP provenance.

Generation quality is competitive with DALL-E 3 inside [ChatGPT](/tools/chatgpt) for realistic photographic images and product visuals. It trails [Midjourney](/tools/midjourney) on aesthetic quality for artistic, painterly, or highly stylised outputs. For commercial photography-style images where accuracy and commercial clearance matter more than aesthetic quality, Firefly is the practical choice. See [Adobe Firefly's current model capabilities](https://www.adobe.com/products/firefly.html) for the full feature list.

Generative fill and Photoshop integration

Photoshop's generative fill is the most-used Firefly feature for working designers. The workflow is straightforward: make a selection with any selection tool, open the Contextual Task Bar, type a description of what you want, and Photoshop generates three variations on a new layer. You accept one, delete the others, and continue editing.

For removal tasks — deleting unwanted objects, people, or backgrounds — generative fill is faster and more accurate than Photoshop's legacy content-aware fill. The AI understands the scene context, so removing a person from a crowded street generates a background that matches the surrounding pavement, architecture, and lighting rather than tiling texture patches.

Generative expand extends the canvas. If you have a product photo sized for Instagram at 1080×1080 but need a version for a billboard at 16:9, generative expand fills in the extended areas with plausible background. The result often needs light retouching at the edges, but it replaces what would otherwise be a substantial manual compositing job.

These tools are available in Photoshop on Creative Cloud subscriptions and consume generative credits. Each plan includes a monthly credit allocation, and additional credits can be purchased. Heavy retouching workflows — an agency processing hundreds of client images per month — will hit credit limits on standard Creative Cloud plans and need the Firefly Premium add-on or a higher-tier Creative Cloud plan with more credits. [Adobe's credit documentation](https://helpx.adobe.com/firefly/using/generative-credits-faq.html) covers current allocations per plan.

Commercial safety and IP clarity

The IP question matters for commercial AI image use in a way it does not for personal or experimental use. When a brand uses AI-generated images in advertising, any copyright claim against the training data could expose the brand to liability. Midjourney, Stable Diffusion, and most other generative models were trained on scraped internet content that includes copyrighted images. Several ongoing lawsuits against AI companies involve exactly this training data question.

Adobe Firefly was designed with this risk in mind. The training set consists of Adobe Stock (a licensed library), Creative Commons licensed content with appropriate permissions, and public domain material. Adobe provides indemnification for commercial use of Firefly outputs — meaning if a third party brings a copyright claim against a Firefly-generated image, Adobe covers the legal exposure up to defined limits.

For individual designers and small creative teams, this distinction rarely matters in practice. But for agencies producing work for regulated industries, large brands with legal teams that review AI use, or any context where the chain of IP ownership needs to be documented, Firefly's training provenance is a genuine differentiator.

Microsoft Designer and Google's image generation tools offer similar assurances, but neither is as deeply integrated into professional design workflows as Firefly inside Photoshop and Illustrator. For creative professionals who need both quality editing tools and commercial IP clarity, Firefly inside Creative Cloud is the clearest option. A comparison of [Midjourney vs ChatGPT](/compare/midjourney-vs-chatgpt) covers the DALL-E 3 alternative if you want to evaluate the ChatGPT image generation path alongside Firefly.

Adobe Firefly vs Midjourney

Firefly and [Midjourney](/tools/midjourney) serve different primary use cases, which is why many professional creatives use both.

Midjourney is better for generating new images from scratch where aesthetic quality is the priority. The colour richness, compositional intentionality, and visual coherence of Midjourney outputs are higher than Firefly's for creative, artistic, and stylised work. If you are generating campaign hero images, editorial illustrations, or mood boards for art direction, Midjourney produces better-looking images.

Firefly is better for editing existing images and for any work that requires commercial IP clarity. Photoshop's generative fill, background removal, and canvas extension are faster and more reliable in the Firefly-powered Photoshop workflow than in any Midjourney-adjacent workflow that requires exporting, editing in a separate tool, and re-importing.

The practical setup for many professional designers: generate conceptual imagery and mood references in [Midjourney](/tools/midjourney), bring the approved direction into Photoshop, and use Firefly for all the editing, retouching, and production work. The combined cost — Midjourney Standard at $30/month plus Creative Cloud (which includes Firefly) — is significant, but for professionals doing regular commercial work, both tools are earning their place in the workflow.

Who should not use Adobe Firefly

Do not use Firefly as a standalone image generator if you do not already use Adobe Creative Cloud. The best Firefly features — generative fill, generative expand, Illustrator vector generation — are inside Photoshop and Illustrator. The web-based Firefly interface at firefly.adobe.com generates images, but the output quality is below [Midjourney](/tools/midjourney) and the workflow is less useful without the Photoshop integration.

Avoid Firefly for high-end aesthetic image generation. If you want to produce images that look like they were art-directed by a professional — rich colour, painterly quality, strong visual identity — Midjourney is the better tool. Firefly's generation quality is competent but not the best in the category.

It is not the right choice if your team does not need commercial IP clearance. The training data advantage matters for enterprise and agency use, but for personal projects, educational work, or any context where you are not selling or publishing the images commercially, paying for Creative Cloud primarily to access Firefly is not a justified expense.

And if your workflow is entirely outside Adobe's ecosystem — you use Figma, Sketch, or other design tools — the integration advantage disappears. [ChatGPT](/tools/chatgpt)'s built-in DALL-E 3 or [Midjourney](/tools/midjourney) are more useful in non-Adobe workflows where the Photoshop integration is not accessible.