Loading the page.
Let’s talk
← Back to the blog

AI & design / Practical guide

Adobe Firefly Graph tutorial: build a reusable AI workflow

One product photo. Three campaign directions. Then another launch next week. Adobe Firefly Graph connects creative tasks into a workflow you can revisit. This tutorial starts with a small product-image job and shows what to check before you reuse it.

A blue ceramic vessel beside three physical image proofs showing different framings of the same subject.
One subject explored through a family of image studies. AI-generated editorial concept by DBS.

What is Adobe Firefly Graph?

Firefly Graph is a visual workspace where connected nodes describe a creative process. One node supplies an image or text. Another edits or generates content. An output node gives you the result. You can return to a step and change it without rebuilding the whole workflow.

Adobe highlighted Firefly Graph in beta at Adobe MAX on November 10, 2026. The same announcement included Firefly Builder in private beta. Graph is the workflow canvas covered here; Builder has a separate waitlist.

The interesting part for a small team is repeatability. If every product launch starts with the same background cleanup and creative treatment, recording those steps can make the next brief easier to handle.

Sources: Adobe’s MAX 2026 announcement · Adobe’s Firefly Graph overview

Choose a job small enough to judge

Start with something you can check by eye: preparing a product image for a campaign, comparing a few background treatments or cleaning up a batch of photographs. Pick one approved source file and one clear output.

For example, a skincare brand could use a bottle photograph to explore a restrained campaign background. The bottle shape, label and cap are fixed requirements. The lighting and surrounding scene are the variables. Write that distinction into the brief before opening a tool.

Keep the original photograph beside your result. A convincing image can still change the product in ways a customer would notice.

How to create your first Firefly Graph

If a connection will not attach, check its data type. An image output needs a compatible image input; port colors help you identify what can connect.

  1. Open Graph in Firefly

    Sign in to Adobe Firefly and select Graph in the left panel. Start with Create new graph or customize a template.

  2. Add an input

    Use Add node to choose Input Image or Input Text. Upload your source asset or enter your brief.

  3. Connect an editing step

    Drag from the input’s output port and choose a compatible node. Connect the matching ports, then adjust its settings.

  4. Add an output

    Connect the final editing or generation node to an appropriate output node so you can inspect and download the result.

  5. Run, review and adjust

    Select Run all, inspect the result and change one input at a time. Check the credit indicator before repeating generative steps.

At a glanceA first workflow you can actually review
  1. Approved input

    Start with one source image and one specific brief.

  2. One operation

    Connect the editing or generation step your task needs.

  3. Inspect output

    Compare the result with the source before running more assets.

DBS example workflow. Actual node names and available operations depend on Adobe’s current release.

Adobe’s demonstration shows a product image moving through a repeatable creative workflow. Use it to understand the approach; check the linked setup guide for current controls.

Video: Adobe Firefly Graph Demo: AI Creative Workflows in Action by Adobe for Business.

The YouTube player connects only when you press play. You can also use the video link above.

Sources: Open Adobe Firefly · Adobe’s create-and-share guide

Use AI Assistant to build the first draft

You can describe a workflow to Graph’s AI Assistant and ask it to create or change the nodes. Adobe says the assistant cannot run the graph or add your input assets. You still supply the files, review the connections and start execution.

Give it a specific job, explain what must stay unchanged and ask it to keep the workflow simple. Review the proposed steps before spending credits.

A starting brief for a product campaign

Help me build a workflow for one product photograph. I want to compare three simple background treatments while preserving the product silhouette, label and colors. Show me where to add my image, which settings I can change and how to inspect each output. Explain any generation steps before I run them.

Sources: Adobe’s Graph AI Assistant guide

Example: a product photography workflow for three campaign images

Imagine you are preparing a launch for a fictional skincare serum. You have an approved bottle photograph and need three campaign directions: a clean studio scene, a warm bathroom shelf and a close-up with generous space for a headline.

Keep the source photograph and the product requirements fixed. Change the scene brief between directions, then compare the outputs at the same size. This makes the review about art direction rather than three different versions of the bottle.

Start with one background treatment. If the bottle label, cap or proportions change, repair that result before expanding the workflow. A batch of ten attractive mistakes still leaves the team with ten corrections.

Keep a simple review sheet: source filename, prompt, model or operation, output, corrections needed and approval decision. Record the time spent correcting the result as well as the time spent generating it. That is the more useful basis for deciding whether to reuse the graph.

A practical acceptance check

The bottle matches the approved packshot. The label remains readable. The shadow fits the surface. There is room for editable campaign text. The final image works at mobile size.

At a glanceOne source photograph. Three directions to compare.
  1. Clean studio

    Quiet background, clear silhouette and room for the campaign headline.

  2. Warm setting

    A contextual surface and controlled props, with the same product scale.

  3. Detail-led

    A closer composition that keeps the actual label and finish readable.

DBS example art direction for a fictional serum. This is a planning guide, not a screenshot or a claim of generated results.

Sources: Explore DBS AI Creative Lab services

Keep the brand consistent across every output

A reusable AI creative workflow can repeat a mistake just as efficiently as a good decision. Approve one output before feeding the workflow more assets. Check it at the size a customer will actually see, then zoom in on the product details.

  • Check the product

    Compare the label, proportions, finish and small details with the original photograph. A generated extra button or a misspelled ingredient needs correction.

  • Check the art direction

    Use a small set of approved references for light, composition and color. Decide which parts of the treatment should stay consistent across the campaign.

  • Keep final typography editable

    Add headlines, prices and calls to action in a design layout where your team can proofread and update them. Check spelling, currency and offer dates for each market.

  • Name the approved version

    Save a clear record of the source file, selected settings and approved output. Your colleague should be able to tell which version is ready to use.

Try it with your project

Review before you reuse the workflow

Compare one output against the approved source and the brief.

0 of 5 checked

Your selections stay on this page until you leave or refresh. Copy them to keep a record.

Check access, credits and sharing before scaling up

For teams in the US, UK, Canada and EU, start with the options shown in your own Adobe account. Check the current local plan and credit allowance before choosing a workflow for client production. A beta announcement alone does not confirm access for every account.

Credit use depends on the nodes and models you run. Adobe’s FAQ lists Graph operations that consume credits even when the corresponding feature is unmetered elsewhere. Try one representative asset and check the actual usage before estimating a larger batch.

Graph offers viewer and editor sharing, but Adobe’s guide says it does not support real-time collaborative editing. Agree who is making changes and keep a clearly named copy for review.

Sources: Adobe’s generative credits FAQ · Adobe’s create-and-share guide

Why a Firefly Graph workflow fails, and what to check

  • A cable will not connect

    Check that the output and input accept compatible data types. Adobe uses port colors to help identify them.

  • The workflow produces the wrong image

    Inspect the earliest step where the result diverges from the brief. Compare it with the source before changing later nodes. Keep one variable steady while testing another.

  • The product looks different in every version

    Use one approved reference and a written list of fixed details. Generative results still need review; asking a tool to preserve a label does not prove that it did.

  • Credit use is higher than expected

    Inspect which nodes ran and how often. Adobe’s “Credits vary” indicator means the exact total depends on execution. Test a small input before expanding the job.

At a glanceFind the first step that needs correcting
  1. Input

    Is this the approved photograph and product variant?

  2. Operation

    Did this step alter a fixed product detail?

  3. Output

    Does the selected result meet the brief at its final size?

Review the source and intermediate result before spending another run on the whole workflow.

Sources: Adobe’s node connection instructions · Adobe’s generative credits FAQ

Questions about Firefly Graph workflows

  • What is a node in Firefly Graph?

    A node handles a particular step, such as an input asset, an editing operation or an output. Connections pass information between compatible nodes.

  • Can I compare models?

    Yes. Graph supports passing the same inputs to different models so you can compare results. Keep the brief and source assets consistent when judging the outputs.

  • Will it keep my logo and packaging perfect?

    Check each result against the approved artwork. Inspect the lettering and product details before publishing, particularly if a generative step has touched them.

  • What should my team try first?

    Choose a recurring task with an obvious pass or fail. Run one asset, record the time and credit use, and ask a teammate to review it. Reuse the workflow when the result meets the brief.

  • Is Firefly Graph free to use?

    Check the access and allowance shown in your Adobe account. Generative operations can consume credits; availability of the canvas does not mean every workflow has unlimited free runs.

Sources: Adobe’s Firefly Graph overview

Your next step

Bring us the task you keep repeating.

Share the assets, formats and brand rules behind your next campaign. We can help decide where an AI-assisted workflow fits and where the work needs a designer’s attention.

Explore AI Creative Lab

What happened when you tried?

Which creative task would you put into a reusable workflow? Tell us what you tried, what worked or where you got stuck.

Share your feedback

Share a question, correction or your experience. We can reply to your email.

Email us directly at hello@designerbstudio.com.

Replies go to our inbox; they are not posted publicly. Please leave out verification codes and identity documents.

Written by Designer Bhai Studio

We design brands, build websites and AI products, and help businesses improve their search presence.

Explore our services

Tell us
what’s next.