Generate Variations generate-variations

If you are looking for a way to optimize your digital channels and accelerate content creation, you can use Generate Variations. Generate Variations uses generative Artificial Intelligence (AI) to create content variations based on prompts; these prompts are either provided by Adobe or created, and managed, by users. After creating variations, you can use the content on your website, and also measure their success using the Experimentation functionality of Edge Delivery Services.

You can access Generate Variations from:

You can then:

Legal and Usage Note legal-usage-note

Generative AI and Generate Variations for AEM are powerful tools – but you are responsible for use of the output.

Your inputs to the service should be tied to a context. This context can be your branding materials, website content, data, schemas for such data, templates, or other trusted documents.

You must evaluate the accuracy of any output as appropriate to your use case.

Before using Generate Variations you are required to Agree to the Adobe Generative AI User Guidelines.

Usage of Generate Variations is tied to the consumption of generative actions.

Overview overview

When you open Generate Variations (and expand the left panel) you see:

Get Started get-started

The interface guides you through the process of generating content. After opening the interface, the first step is to select the prompt you want to use.

Select Prompt select-prompt

From the main panel, you can select:

To differentiate:

Provide Inputs provide-inputs

Each prompt needs you to supply certain information so that it is able to get appropriate content back from generative AI.

The input fields guide you through what information is needed. To help, certain fields have default values that you can use, or modify as needed, and descriptions explaining the requirements.

There are several key input fields that are common to multiple prompts (certain fields are not always available):

Generate Copy generate-copy

After you fill out the input fields, and/or modify the prompt, you are ready to generate content and review the responses.

Select Generate to see responses generated by generative AI. The generated content variations are shown under the prompt that generated them.

When you select a single variation, the following actions are available:

If you need to refine the inputs or prompt, you can make adjustments and select Generate again to get a set of new responses. The new prompt and response are shown below the initial prompt and response; you can scroll up and down to view the various sets of content.

Above each set of variations is the prompt that created them, along with a Re-use option. If you ever need to re-run a prompt with its inputs, select Re-use to reload them in Inputs.

Generate Image generate-image

After you have generated text variations, you can generate images in Adobe Express using the Generative AI capabilities of Firefly.

Select a variation, followed by Generate Image, to directly open Text to Image in Adobe Express. The prompt is pre-populated based on your variant selection, and images are automatically generated according to that prompt.

You can make further changes:

You can also use Explore More for further possibilities.

When complete, select the desired image and Save to close Adobe Express. The image is returned, and saved with the variation.

Here you can mouse-over the image to show action items for:

Use Content use-content

To use the content generated with generative AI, you must copy the content to the clipboard for use elsewhere.

This is done using the copy icons:

Once copied to the clipboard, you can paste the information for use when authoring content for your website. You can also run an experiment.

Favorites favorites

After reviewing the content you can save selected variations as favorites.

Once saved they are shown under Favorites in the left navigation. Favorites are persisted (until you Delete them, or clear the browser cache).

Recents recents

This section provides links to your recent activity. A Recent entry is added after you select Generate. It has the name of the prompt and a timestamp. If you select a link, it loads the prompt, populates the input fields as appropriate, and shows generated variations.

Edit the Prompt edit-the-prompt

The underlying prompt can be edited. You may want to do this:

Select Edit Prompt:

This opens the prompt editor, where you can make your changes:

Add Prompt Inputs add-prompt-inputs

When creating or editing a prompt, you may want to add input fields. Input fields act as variables in the prompt and give the flexibility to use the same prompt in various scenarios. They allow users to define specific elements of the prompt, without having to write the entire prompt.

Example: Add new text field - Tone of Voice example-add-new-text-field-tone-of-voice

To add a new text field titled Tone of Voice, use the following syntax in your prompt:

{{@tone_of_voice,
  label="Tone of voice",
  description="Indicate the desired tone of voice",
  default="optimistic, smart, engaging, human, and creative",
  type=text
}}

Create a prompt create-prompt

When you select New prompt from Prompt Templates, a new panel will allow you to input a new prompt. You can then specify these, together with the Temperature, to Generate content.

See Save Prompt for details of saving the prompt for future.

See Add prompt inputs for details about adding your own prompt inputs.

If you would like to preserve formatting both in the UI, and when copied and pasted into the document based authoring flow, include the following in the prompt:

The following image shows the advantages of doing so:

Save Prompt save-prompt

After editing or creating prompts, you may want to save them for future use; for either your IMS organization or yourself alone. The saved prompt will appear as a Prompt Template card.

When you have edited the prompt, the Save option is available at the bottom of the Inputs section, to the left of Generate.

When selected, the Save Prompt dialog opens:

Audiences audiences

To generate personalized content the generative AI must have an understanding of the audience. Adobe provides a number of default audiences, or you can add your own.

When adding an audience, you should describe the audience in natural language. For example:

Two Audience Sources are supported:

Audience - Adobe Target audience-adobe-target

Selecting an Adobe Target audience in the prompt allows for the generation of content to be personalized to that audience.

Add Adobe Target Audience add-adobe-target-audience

See Create audiences to create an audience in Adobe Target.

Audience - CSV file audience-csv-file

Selecting a CSV file audience in the prompt allows for the generation of content to be personalized to the selected Target Audience.

Adobe provides a number of audiences to use.

Add Audience CSV file add-audience-csv-file

You can add a CSV file from various platforms (for example, Google Drive, DropBox, Sharepoint) that have the capability to provide a URL to the file once it is made publicly available.

For example, to add an audience from a file on Google Drive:

Generative Action Usage generative-action-usage

Usage management depends on the action taken:

Access Generate Variations access-generate-variations

After fulfilling the prerequisites you can access Generate Variations from AEM as a Cloud Service or the Sidekick of the Edge Delivery Services.

Access Prerequisites access-prerequisites

To use Generate Variations you must ensure that the prerequisites are fulfilled:

Access to Experience Manager as a Cloud Service with Edge Delivery Services access-to-aemaacs-with-edge-delivery-services

Users who need access to Generate Variations must be entitled to an Experience Manager as a Cloud Service environment with Edge Delivery Services.

To grant access to specific users, assign their user account to the respective product profile. See Assigning AEM Product Profiles for further details.

Access from AEM as a Cloud Service access-aemaacs

Generate Variations can be accessed from the Navigation Panel of AEM as a Cloud Service:

Access from the AEM Sidekick access-aem-sidekick

Some configuration is needed before you can access Generate Variations from the Sidekick (of Edge Delivery Services).

Further Information further-information

For further information you can also read:

FAQs faqs

Formatted Output formatted-outpu

The generated response isn’t giving me the formatted output that I need. How do I modify the format? ex: I need a title and a sub-title but, the response is just title

Length of Response length-of-response

The generated response is too long or too short. How do I change the length?

Improve responses improve-responses

The responses I’m getting aren’t exactly what I’m looking for. What can I do to improve them?

Comments in a prompt comments-in-prompt

How can I use comments in a prompt?

Comments in a prompt are used to include notes, explanations, or instructions that are not meant to be part of the actual output. These comments are encapsulated within a specific syntax: they start and end with double curly braces and begin with a hash (for example, {{# Comment Here }}). Comments help clarify the prompt structure or intent without affecting the generated response.

Find a shared prompt find-a-shared-prompt

What can I do if I cannot find a prompt template that someone has shared?

In this situation there are various details to check:

Custom prompts in v2.0.0 custom-prompts-v200

In v.2.0.0 my custom prompts have disappeared - what can I do?

Moving to the v2.0.0 release will cause custom prompt templates to break - so they will be unavailable.

To retrieve them:

Release History release-history

For details of the current, and previous, releases see the Release Notes for Generate Variations