Media in the age of AI generation
AI image generation, AI voice synthesis, and AI scenario writing are now standard options in the eLearning production toolkit. They reduce costs, compress timelines, and remove dependencies on external photographers, voice artists, and script writers for standard assets.
They also introduce new briefing requirements. A human creative professional will ask clarifying questions. An AI generation tool will not — it will produce something, confidently, based on exactly what you typed. The quality of the brief is the quality of the asset.
By the end of this unit- Write an AI-ready image brief that produces eLearning-appropriate output.
- Identify when AI voice generation is suitable and when human narration is required.
- Use AI to generate scenario dialogue that is realistic and instructionally grounded.
The same principle, a different medium
Everything you have learned about prompting for written ID content applies to media briefing — with additional constraints specific to each medium. The five-element prompt anatomy (role, audience, task, format, constraints) still applies. What changes is the nature of the constraints and the specific quality checks required for visual and audio assets.
Image generation for eLearning
AI image generation is most useful in eLearning for three use cases: contextual illustrations (showing a workplace setting or situation), diagram replacements (abstract visual representations of concepts), and custom stock photography alternatives. It is not currently reliable for interface screenshots — use actual software captures for those.
Writing an AI image brief
An AI image prompt for eLearning differs from a creative prompt in one important respect: it needs to produce a specific, instructionally appropriate image, not a visually striking one. The brief must prioritise accuracy over aesthetics.
Required brief elements for eLearning images
Scene description: What the image shows, described precisely. Who is in it, what they are doing, where they are. Not "a professional in an office" — "a mid-career woman reviewing a laptop screen in a modern open-plan office. No other people visible. Daytime. Neutral background."
Style specification: Realistic photography, flat illustration, diagram, or infographic. State this explicitly. AI defaults to a style based on word patterns in your prompt — if you want flat illustration but describe a realistic scene, you may get a photograph.
Inclusions and exclusions: State what must be in the image (the laptop must show a visible screen, the person must be alone) and what must not (no text overlaid, no phones, no obvious AI imagery like robot hands or glowing circuits).
Accessibility note: Write the alt text in the brief, before the image is generated. This forces specificity about what the image needs to communicate and ensures accessibility is designed in, not retrofitted.
Common quality failures in AI-generated eLearning images
- Hands and text: AI image generators frequently produce incorrect hand shapes and illegible or nonsensical text. Always zoom in on hands and any readable text in the image before using it.
- Generic stock aesthetics: Without explicit direction, AI defaults to the visual language of stock photography — extreme diversity clichés, unnaturally bright lighting, exaggerated expressions. Brief against this explicitly.
- Misrepresented technology: AI-generated images of software interfaces are usually wrong. Use real screen captures for any interface that learners need to recognise.
- Inconsistency across a module: AI-generated images do not maintain visual consistency between generations. For character continuity across a module, either use a single generation session with consistent descriptions or use human-created illustrations.
For contextual illustrations that do not require character continuity — a workplace setting, an abstract concept, a process diagram — AI image generation is highly effective. For scenarios requiring the same character to appear across multiple screens, plan for human illustration or consistent character generation within a single AI session.
AI voice and narration
AI voice synthesis has reached a quality level that makes it suitable for most workplace learning narration. The decision between AI voice and human narration is now a design decision, not a budget shortcut — each has appropriate use cases.
When to use AI voice, when to use human narration
- Standard module narration for process-led or information-heavy content
- Content that will require frequent updates (AI re-generation is faster and cheaper than re-recording)
- Multi-language versions where consistent voice across languages matters
- Accessible content (captions and transcripts are generated automatically by most AI voice tools)
- Quick turnaround projects with limited budget
- Senior leadership messages where the speaker's identity and voice carry the message
- Emotionally complex content — mental health, safety incidents, sensitive change communication
- Content where a specific known person's voice is part of the learning experience (a subject matter expert, a founder)
- Content where learner trust in the messenger is a design requirement
Briefing AI voice generation
The voice generation brief has three components: the script (verbatim, as it will be spoken), the voice parameters (tone, pacing, gender, accent where the tool allows), and the output specification (format, sampling rate, whether pauses are needed at specific points).
Review a piece of narrated eLearning you have designed or evaluated recently. Would AI voice generation have been appropriate for that content? What would have been gained, and what (if anything) would have been lost?
Scenario and dialogue generation
AI is particularly useful for generating scenario text and character dialogue at volume — creating multiple versions of the same scenario for different roles, generating branching dialogue trees, and producing realistic workplace conversations that would take significant time to write from scratch.
Briefing for scenario generation
Anchor the scenario in a real role and task
The most important element of the scenario brief is specificity about the role and the real work task. "A procurement manager reviewing supplier contract terms" produces more useful content than "an employee dealing with a work situation." Generic role descriptions produce generic scenarios.
Specify the decision point, not the resolution
Brief AI to produce the scenario up to and including the decision point — not through to the resolution. You will design the resolution (the branching options and their consequences). Ask the AI for the situation and the dilemma; write the choices yourself.
Brief for character constraints, not just character descriptions
Tell the AI what constraints the characters are operating under: time pressure, incomplete information, conflicting priorities. Unconstrained characters make perfect decisions. Constrained characters make the kind of decisions your learners actually face.
Generate multiple versions and select
Ask Copilot for three or four versions of the same scenario with different emphasis or different character perspectives. Choose the version that most closely matches your learner's real context, then edit. Selecting from multiple options is faster than iterating a single output.
Which of these scenario briefs will produce the most instructionally useful output from Copilot?
End of Unit 7
You should now be able to:
- Write an AI image brief that includes scene, style, inclusions/exclusions, and alt text.
- Select between AI voice and human narration based on content type and design requirements.
- Brief Copilot to generate scenario text anchored in a specific role, with a defined decision point and character constraints.
Unit review
Which type of eLearning image should NOT be AI-generated?
Which type of content most requires human narration rather than AI voice?
Why should you brief Copilot to stop a scenario at the decision point rather than including the resolution?
What is the most important quality check for an AI-generated image before it is used in a module?
End of module
You have completed Course 07: Briefing AI for Learning Media. Next: Evaluating AI-Augmented Learning — measuring whether AI-focused learning actually changes behaviour, and using agents to automate evaluation.