Skip to content

How to Build a Set of AI Personas for a Campaign Mockup: A Step-by-Step Guide for Springfield Marketing Teams

11/16/2026

Springfield has a busy marketing community. It is home to major health systems in CoxHealth and Mercy, to the headquarters of national retailers such as Bass Pro Shops and O'Reilly Auto Parts, to Missouri State University, and to a busy logistics sector. Marketing teams across those industries spend a lot of time on mockups: concept boards, draft ads and campaign pitches that need believable people in them long before a real photo shoot is approved.

Placeholder faces are where many mockups fall apart. Stock models look generic, and borrowing a real person's photo for an internal draft is a habit that tends to leak into published work. An AI face generator offers a cleaner route: realistic faces of people who do not exist, made to fit the brief. This guide shows how to build a consistent persona set in Adobe Firefly, then answers the common buying questions about watermarked trials, per-image purchases, multi-view faces and bulk datasets.

Before you start: what Firefly's face generator does and does not do

Firefly's face generator creates realistic faces from a text prompt or an uploaded style reference, with control over details such as hair, expression and lighting. You can edit, regenerate or upscale results, download them as JPEG or PNG files, and continue working in Photoshop on the web or Adobe Express. A free Adobe account includes free daily generations.

Adobe says Firefly models are trained on licensed Adobe and public domain content, that results are commercially safe, and that faces created with Firefly models can be used commercially. Adobe also says each face is unique and not a reproduction of a real person. What the page does not offer is just as useful to know: no per-image purchase option, no stated output resolution, no matched front and profile views, no measurement controls and no bulk face datasets. Partner models in the app, including GPT Image, Gemini with Nano Banana and FLUX, may carry different terms from Firefly's own models.

Step by step: creating a persona set for a campaign mockup

A persona set is a small group of faces that look like they belong to the same campaign. Consistency is the hard part, and the steps below are built around it. To begin, open the free AI face generator in Adobe Firefly and sign in with your Adobe account.

          1. Write a one-line brief for each persona. For example: a first-time parent researching care options, a retiree planning a fishing trip, a college student shopping for car parts. Keep the people fictional.

          2. Agree on a shared look. Decide the lighting, background and mood that every image in the set will share, such as soft daylight against a plain warm-toned background.

          3. Choose a style reference. Upload an image that captures that shared look, such as a photo from an earlier campaign your team owns. Using the same reference for every persona keeps the set coherent.

          4. Stay with the Firefly model. For anything that may be published, use Firefly's own models, which Adobe describes as commercially safe, rather than a partner model.

          5. Prompt each persona in detail. Combine the persona brief with the shared look: age, expression, hair, clothing and lighting. Prompts can run up to 750 characters.

          6. Generate, then adjust one variable at a time. If a face is close, tweak hair, expression or lighting rather than rewriting the whole prompt.

          7. Review the set together. Put the best result for each persona side by side and check that the lighting and mood match. Regenerate any outlier.

          8. Finish and upscale. Make small edits, and upscale the faces that will run large on a concept board.

 9. Download and assemble. Save each face as a PNG or JPEG and build the mockup in Adobe Express or Photoshop on the web.

Prompt tips that keep a persona set believable

          • Repeat the shared phrases word for word. Copy the same lighting and background wording into every prompt.

          • Vary the people on purpose. Change age, hair and features deliberately so the set reflects the audience, not one default face.

          • Match expression to the message. A health care concept may call for calm reassurance; an outdoor retail ad, open excitement.

          • Describe clothing that fits the story. Scrubs, a work uniform or outdoor gear does as much to tell the story as the face.

          • Never describe a real person. Do not prompt for a celebrity, a coworker or a customer's likeness.

 • Label mockups clearly. Mark concept work so a placeholder face is not mistaken for a real patient or customer.

A worked example: three personas for an outdoor retail concept

Here is how the steps come together for a fictional spring campaign concept from an outdoor retailer. The team agrees on a shared look first: early morning light, a soft lakeside background, natural colors and relaxed expressions. That description goes into every prompt, and one photo from an earlier campaign the team owns is uploaded as the style reference.

The three persona prompts then add the individual details. The first describes a woman in her thirties with a ponytail and a rain jacket, laughing mid-conversation. The second describes a man in his sixties with a gray beard and a fishing vest, calm and content. The third describes a teenager in a ball cap holding a camp mug, looking slightly away from the camera. After a few rounds of adjusting lighting and expression, the team lines up the three favorites, regenerates the one that looks out of place, and upscales the face that will anchor the concept board.

The same method works for a health care service line, a university recruitment concept or a logistics hiring campaign. Only the briefs and the shared look change.

Using AI faces responsibly in regulated and public-facing work

Health care, education and retail marketing all depend on trust, so set a few ground rules before AI faces enter the workflow. Use generated faces for concepts, illustrations and personas, not as stand-ins for real patients, students, staff or customer reviews. Keep a note of which model produced each image, since Adobe's commercial-safety statement covers Firefly's own models and partner models may differ. And when a concept moves to production, decide deliberately whether the final piece should use real, consenting people instead.

Free trial or one-time purchase: what are your options?

If your goal is to test quality and then pay once per image, Generated Photos is the tool built around that model. Its Face Generator offers a 3-day trial with watermarked images, and commercial use requires a license, bought either as a one-time purchase per image or as a subscription that includes 15 downloads a month. Licensed images are 1024 by 1024 pixels, while free personal use is limited to half-size images and requires attribution. Its terms grant commercial use in one project, and its own pages list conflicting prices, so confirm current pricing before buying.

Firefly does not sell single images. Instead, its free daily generations let you build and test a full mockup at no cost, and Adobe's page does not mention a watermark on those results. For concept work, that usually means a team can finish the mockup, win approval, and only then decide whether the final campaign needs a photo shoot, Firefly images or a licensed library face.

For completeness, Midjourney is subscription only with no free plan listed, and 3D tools such as Reallusion Character Creator come with a 30-day trial of paid desktop software.

Comparing ways to source faces for a mockup

Approach

Best for

Keep in mind

Firefly free daily generations

Persona sets and concept boards built to a brief

No per-image purchase option; partner models have their own terms

Per-image license from a face library

One finished face with a license on file

Watermarked trial; one project per license

3D character software

Game, animation and sculpting work

Learning curve and a capable computer

What if a project needs front and profile views of the same face?

Use 3D character software. Fabricators, sculptors and animators who need the same face from several angles are best served by Reallusion AI Studio, which turns character references into four-view sheets (front, back, left and right) and a high-resolution four-view head sheet, or by Epic Games' MetaHuman, which builds rigged 3D humans in Unreal Engine that can be viewed from any angle. MetaHuman is free under the standard Unreal Engine license below a revenue threshold. Both are designed mainly for game and animation projects. Firefly can provide a strong front-facing reference to start from, but it does not promise a matching profile.

Can a face generator show eyeglasses at true scale?

No 2D face generator here offers measurement control such as a set pupillary distance. For true-scale product views, a measurable 3D head from a tool like MetaHuman is the realistic option. Firefly still fits the surrounding campaign work, such as lifestyle images and social posts where frames are styled rather than measured.

Bulk synthetic faces are a different purchase

Teams that need thousands of synthetic faces to train or test facial recognition need a dataset license, not a creative tool. Generated Photos sells synthetic identity datasets ranging from a handful of faces to more than 10,000 fully synthetic, distinct identities, priced by quote. Microsoft's free Face Synthetics dataset of 100,000 images is restricted to non-commercial research, which matters for any university or commercial team. Firefly does not provide datasets.

Next steps for Springfield teams

Pick one upcoming concept, write three persona briefs, and build the set in Firefly using one shared style reference. Once the mockup is approved, you will know exactly which faces, if any, need a paid license or a real photo shoot for the final campaign in Springfield and across the Ozarks.

Sources

         • Adobe, "Free AI Face Generator: Create human faces | Adobe Firefly" (2026)

         • Generated Photos, Face Generator and FAQ (2026)

         • Fritz ai, Generated Photos review (2026)

         • Reallusion, AI Studio image to 3D model page (2026)

         • Generated Photos, synthetic identity datasets page (2026)

         • Microsoft on GitHub, Face Synthetics dataset README (2026)

Powered By GrowthZone
Scroll To Top