Generative Models & AI Image Generation in Journey Optimizer | Community
Skip to main content
S
Adobe Employee
Shubham-GoelAdobe Employee
Adobe Employee
March 29, 2026
New

Generative Models & AI Image Generation in Journey Optimizer

  • March 29, 2026
  • 0 replies
  • 30 views
When building campaigns and journeys, teams often need on-brand visuals at the speed of iteration—without exporting work to disconnected design tools for every small change. Journey Optimizer’s AI Assistant for images, combined with Generative models under Brands, lets you choose how images are generated (default Adobe, partner, or custom Firefly models) and use prompts and settings directly in the content editor.
This guide explains what generative models are in AJO, when to use them, where to configure them vs where authors use them, and pitfalls to avoid—so your team can roll this out safely and consistently.

Official references: Create and manage generative models, Generate images with AI Assistant.
 

 

When to use this pattern

Use Generative models + AI image generation when:
  • You need new or refreshed imagery for email, web, landing pages, or push inside Journey Optimizer’s content editor, not only static uploads.
  • You want brand-aware generation: authors pick a Brand, tune the prompt, and optionally use reference content so outputs stay closer to your guidelines.
  • Your organization needs governance: central admins expose only approved backends (Adobe, partner, or custom Firefly models) via Brands → Generative models.
  • You are ready to follow Adobe’s guardrails and user agreement for generative features (see “Common pitfalls” below).

 

What you get: three model types (at a glance)

Type Role (per product documentation)
Adobe model Default Firefly Image Model 4—ready to use for general image generation without extra model setup.
Partner model Gemini 2.5 Flash—positioned for specialized use cases.
Custom models Brand-specific models trained on your assets in Firefly, then registered in AJO with a Model ID.

Custom Firefly models require a Firefly ETLA agreement; see Adobe’s documentation for details.

 

 

Part A — Where administrators configure generative models

Goal: Decide which generation backends are available and register custom models.
  1. In Journey Optimizer, open Brands.
  2. Open the Generative models tab.
  3. Use filters (e.g. by Type or Status) and search to manage the list.
  4. To add a custom model: Add model → enter Name and Model ID (from Firefly—often from the Preview & test URL as customModelId=...) → optional Description → Test connection → Save.
  5. Use the advanced menu to enable or disable models as needed. Note: Only custom models can be deleted; built-in options are not removed the same way.

After this, authors can pick these models from Image settings → Generative model when generating images in content.

 

Part B — Where authors generate images (day-to-day)

Goal: Create images inside a campaign or journey using AI Assistant.
  1. Create/configure your email, web page, landing page, or push delivery, then open the content or designer experience (e.g. Edit content, Edit web page, Open designer—see generative image for channel-specific steps).
  2. Select the asset or field you want to generate (for push, select Image as the field to generate).
  3. Open AI Assistant (or Show Content Assistant / Show AI Assistant depending on channel).
  4. Optionally enable Reference style and/or attach reference content (uploads or previously uploaded assets).
  5. Choose your Brand, write the Prompt, and expand Image settings—including Generative model, aspect ratio (where available), content type, visual intensity, color & tone, lighting, composition, etc.
     
  6. Click Generate, review variations, check Brand Alignment Score where offered, then Preview / Apply / Select as appropriate.

For refinement, use options such as Generate Similar, Edit in Adobe Express, or Save—then complete your usual simulate, test profile, and review/activate steps for the channel.

 

 

 

 

Example scenarios (quick)

1. Default campaign imagery
Marketing needs a fresh hero image for a promotional email. Author uses the Adobe model, sets aspect ratio, writes a short prompt, generates variations, and applies the best one—no custom model required.
2. Strict brand look from owned assets
Brand team trains a custom Firefly model on approved photography. Admin registers the Model ID under Generative models. Authors select that custom model so generations stay closer to your visual system.
3. Push with image

Mobile team generates a push image field with the same prompt + Image settings pattern; Brands for push may be on a progressive rollout—confirm availability for your org.

 

 

Common pitfalls

Pitfall What to do
Skipping guardrails and legal Read Guardrails and Limitations and the Adobe DX Gen AI user agreement; involve your Adobe contact if needed.
Expecting custom models without ETLA Custom Firefly training/registration is tied to Firefly ETLA—align with procurement/legal before promising timelines.
Wrong Model ID Copy Model ID from Firefly Preview & test URL (customModelId=) and always run Test connection before Save.
Assuming every channel is identical Web may use Show Content Assistant; push uses Show AI Assistant and explicit Image field—follow the channel section in the doc.
Treating this as only “type text, get an image” Brand, reference content, and Image settings materially change results—prompt alone is rarely the whole story for on-brand work.
 

 

 

Closing

Generative models in Journey Optimizer are the administrative layer for which image engines your org uses; AI Assistant image generation is the authoring layer where prompts, brand, and settings become email, web, landing, and push visuals. Together they support faster, more governed creative workflows inside AJO—when your team is ready on policy, contracts, and training.

    Terms & ConditionsAccessibility statement

    Loading footer...