For this recap, I’ll share one or two key points or useful takeaways from each session

Tuesday: Storyline & Rise Pro Lab

AI Assistant

The day started with David Anderson giving an overview of the AI Assistant features in Rise and Storyline. I admit that I haven’t tested out these features too much yet. One feature I noted that I want to test is using the AI to sync objects to audio with AI. That can be a really tedious process, and I’m interested to see how much time that can save me.

Speed Up Development

I also attended Diane Elkins’ session on Tips and Tricks to Speed Up Development. While a lot of these tips were things I already do, I loved her tip about rethinking of Feedback Masters as Layer Masters. While Slide Masters obviously save a lot of design rework, they only work on the base layer. But Feedback Masters are effectively a reusable layout for any layer. We normally only use them for feedback layers, but they could be used for any sort of layer: character dialog, additional content, etc.

Code Block Showdown

The Code Block Showdown was a lot of fun. Jeff Batt gave 4 people (Destery Hildenbrand, Melissa Milloway, Casey Shiray, and Diane Elkins) the same starting request and 5 minutes to vibe code an interactive block for Rise. Then, they each explained their blocks, and everyone in the audience voted on their favorites. It was a great example of what could be done quickly as well as interesting to compare how different their results sometimes were.

My session: Branch Smarter, Not Harder: Naming Conventions, Variables & Bottlenecks

You can see the demo branching scenario I used for my session and how I built it in my previous post about ChronoVoyage Tours. I was overly optimistic about how much interactive practice we’d be able to accomplish in a 40-minute session, so I had more planned than we had time for. Although my timing was off, I’m glad I helped multiple people create their very first variables and triggers in Storyline, and they have the demo files to practice with later.

Wednesday: DevLearn Day 1

The Courage to Think Differently

After the preconference, Day 1 started out with Dan Pontefract’s keynote. One idea I got from the keynote was his idea of a “career canvas” rather than the traditional metaphor of a “career ladder.” Careers aren’t a linear upward climb to successively more valuable roles. Fulfilling careers can be multidirectional.

My Session: Cohesive By Design: Generate Consistent AI Icons, Illustrations, & Characters

I always love leading BYOD (Bring Your Own Device) sessions because people get to apply what they’re learning right away. Check out just a few of the custom icon sets below that participants generated live during my session with just a few minutes of practice! We talked about different styles of icons including isometric, paper cutout, and line icons, plus prompting to keep a consistent color scheme. Participants were able to generate a wide range of styles for icons and illustrations plus a few images of a single character in different poses and scenes.

Gamification that Works

This session demoed some call center scenario training. One of the things I appreciated seeing was their dashboard that showed the stats for the scenario, including a breakdown of responses per question. They also included an AI analysis of the data that summarizes the data, highlights strengths and weaknesses, and recommends next steps. Stakeholders in the business probably don’t really care about the answers to each question (that data is more for the L&D team), but having that summary easily accessible makes it easier to translate that data for them.

Developing AI-Ready Learners

While I don’t work much with higher ed clients, Adam Hogan’s session on human judgment and ethics was worthwhile. He said, “Tools don’t determine outcomes. Human capabilities do.” He noted that our feelings and confidence in AI are often wrong. We need more confidence in ourselves.

You have to calibrate your work with AI. What do you do with the results AI gives you? If you trust it too much, you get errors. If you trust it too little, you don’t get the benefits. The jagged frontier of AI makes it hard to know what AI is good at or not.

Learning happens in the effort: explaining, organizing, connecting, evaluating. Yes, AI can do that work, but if you offload it all to AI, you don’t learn as well.

Should you offload work to AI or stay engaged with it? If any of these are false, stay engaged.

• Low stakes
• Easy to check
• Not the skill you’re trying to build

The Language of Learning: Mastering Writing for Instructional Design

I also got to promote the new book The Language of Learning. Hadiya Nurridin is the one who really deserves all the credit as the editor who kept all of us contributors on track and shepherded us through the whole process. But I’m proud to have contributed a chapter on scripting, and it was exciting to see the physical book for the first time at the conference book store. You can preorder the book now.

Buzz: Video & Media

The Buzz sessions were in a different format this year, with multiple speakers from a single track available at the end of the day to answer additional questions. Samuel Apata and I facilitated a conversation about AI, images, and media. I wish the Buzz sessions had been a little longer; we had plenty of discussion left when we hit the 30-minute mark.

Thursday: DevLearn Day 2

Learning Design in the Age of AI

Luke Hobson started out day 2 of DevLearn with some good laughs. I appreciated his energy and his questions for the panel who joined him for the keynote: Karl Kapp, Tessa Forshaw, and Brandy Ford.

Karl Kapp noted that we as instructional designers add friction, and that’s our value. We ask questions and identify the steps SMEs skip when they explain processes. IDs have lots of valuable skills that help us be more effective working with AI, including our ability to behaviorally define skills, outcomes, examples & non-examples.

Brandy Ford noted that AI is a lever and an enabler, not a replacement. Balance the pressure to adopt AI with solving a real business problem.

Tessa Forshaw discussed the importance of “effortful practice” in the neurobiological event of learning. If we stop doing something, we gradually lose our skills.

Clone Your Best Evaluator: Build AI Rubrics That Actually Capture Quality

While I enjoy hearing from certain speakers over and over at conferences, I also try to seek out some voices I haven’t heard before. Dakota Duffy was one of those new-to-me speakers, and I’m really glad I attended her session on rubrics. This was a 90-minute session, which meant we had some more time to practice. While she was talking about this in terms of giving AI rubrics to evaluate human performance at scale, I think this all could apply to creating rubrics to evaluate AI outputs as well (including having an AI evaluate itself or another AI).

Interview your SMEs for observable evidence you can include in a rubric, not just “the right way.” (These questions work for gathering info for scenarios too!)

• Compare: Which is stronger and why?
• Probe: Which warning sign changes your decision?
• Challenge: Which different approaches are acceptable?
• Boundary: What looks plausible but creates real risk?

The AI doesn’t know what you meant. If you don’t specify, the model infers what counts. Break down larger decisions into separate, more granular decisions.

Four types of scoring ambiguity:

1. Interpretation: What does “appropriate” mean?
2. Sufficiency: How much evidence is enough?
3. Boundaries: When does level 1 become level 2?
4. Construction: Does this show judgment or just fluent writing?

Ditch the Quiz: AI Scenarios

This session had a serendipitous connection to the rubric session I attended before it because they used a rubric for behavior mapping to evaluate responses in their AI roleplays. The speakers used Yoodli as their platform, but there are several other AI roleplay tools available.

The presenters shared how they combined their scenario blueprint (the roleplay structure with context, goal, and personas) with the behavior rubric (behaviors, observable evidence, and feedback).

Instructional Design Buzz

One of the topics of discussion was about simulations and the cautions for working with simulations. They need to feel real. If a simulation feels fake at all, or participants think “nobody talks like that,” it immediately undermines the whole activity. The choices need to be not obvious for a simulation to be valuable.

Another point raised: don’t collect data you don’t need. When you’re deciding what data to collect, ask: What will that data let you do?

Friday: DevLearn Day 3

From Passive to Playable: Interactive Video

I could only attend one session on Friday morning before heading to the airport, and I was looking forward to Buddy Butler’s session on interactive video. One of the key themes of this session was that interactivity in video doesn’t necessarily have to mean branching scenarios or role plays. We can give learners agency at meaningful moments in other ways too, and you can combine different types of interactivity into a single learner experience.

What does the learner need to control?

• Decision point: What should I do?
• Attention: What should I notice?
• Support: What help do I need?
• Flow: Where should I go next?

Conversations and connections

And of course, one of the big reasons you go to a conference like this is to see people in person that you’ve maybe only met online, to have conversations and make new connections. I enjoyed a “dine around” dinner with Dan Keckan and the Cinecraft team plus Cyndi Nagel and Jennifer Fitzpatrick, several receptions and happy hours, and a dinner with Karl Kapp, Jessica Briskin, and the students and alumni of Commonwealth University. That’s in addition to the conversations in the Expo hall and between sessions. As much as I love online, being physically together in the same space has a different energy.

The post DevLearn 2026 Recap appeared first on Experiencing Elearning.

Try the ChronoVoyage Tours scenario

Email readers, if the scenario isn’t embedded below, try opening it in a new tab.

My process with Claude and Flora

I used Claude Opus 5.5 to develop this scenario, including drafting the text and the structure. All of the images were generated in Flora, but most of the image prompts were also written by Claude rather than by me. Then, I built everything in Storyline, adjusting some of the text and adding a few additional images.

Here’s the process step-by-step in more detail, including which tool I used and what portions were my work versus the AI.

Images in Flora

On the Flora canvas, you can see the original three look and feel concepts plus the additional backgrounds in the style I selected.

I used the built-in tool for background removal within Flora to put the characters and objects on a transparent background.

Storyline build

The branching structure in this scenario is deliberately not terribly complicated since it’s designed for others to use for practice. The scenario branches out a little with parallel tracks for Decision 2 but then consolidates to a bottleneck for Decision 3.

Articulate has an MCP that can be connected to Claude, but it doesn’t do this kind of Storyline development work (at least not yet). It would be interesting to see if Claude could follow its own build sheet to develop something in Storyline.

Reflections

Using Claude and Flora definitely sped up the time required for planning, drafting, and image generation. I might do a little more manual prompting for the initial reference images for a real client project, but using Claude to generate the prompts for Flora worked well. This is one of the first times I’ve truly used Claude as an agent to take actions in another program, and I’m happy with the results.

The wrong answer in the scenario are too obviously wrong, even after I tweaked the wording to make the distractors more tempting and plausible. That’s one of the ongoing stumbling blocks I see with AI-written branching scenarios and some dynamically generated scenarios. They’re consistently too easy.

In total, this took about 9 hours for me to build. Without AI, it definitely would have taken me longer to assemble the images from stock illustrations and cutout character images. I would have spent more time on the writing too, but if this was a real client project I’d also need to spend more time writing and revising the text.

But overall, this worked well for my purposes for a Storyline development practice scenario. The AI tools have come a long way since I used ChatGPT to write a prototype scenario back in September 2023.

More on scenarios and AI

Read more about branching scenarios and how I’m using AI in my work.

• Managing the Complexity of Branching Scenarios
• Branch and Bottleneck Scenario Structure
• Branching Scenarios in Storyline: Layers or Slides?
• Speed Up Character Image Creation with Flora
• Updating Scenario Character Images with Flora
• Updating A Twine Branching Scenario with AI – Part 1

The post ChronoVoyage Tours Branching Scenario Example appeared first on Experiencing Elearning.

Video and animation

How long should training videos be?

Rather than specifying a single ideal length for all videos, this article argues for different ranges depending on the purpose and goal. While I’ve cited and used the 2014 research showing a drop in video engagement after 6 minutes, that single number never quite seemed right for all purposes. This article uses more recent research and provides a more nuanced answer.

Lemo-Opuscar · 43 film styles for Claude Code, made in code

A collection of prompts and skills for describing animation styles that Claude can use to generate videos.

Explainer animation Claude skill

This is a Claude Opus 5.5 skill that generates a 30-60 second explainer video based on the info and files in a project. I haven’t tried this myself, but it will be interesting to see how this compares to creating work with more manual direction (even with AI assistance or production). Looking at the skill, it should be possible to describe other styles than the paper craft style here to customize the look and feel.

AI environmental impact

AI Footprint Calculator

I continue to review and bookmark tools that calculate the energy, water, and resource use of AI. I obviously use AI, but considering the energy use is part of how we exercise judgment in when and how we use or avoid AI. This calculator includes more granular AI tasks, separating simple text from deep reasoning and file analysis. It also includes image, audio, and video generation, which many other calculators don’t include. It doesn’t include the comparisons to streaming though, which I think is helpful for perspective.

AI and environmental sustainability

This is the July 2026 update to an older research summary I bookmarked previously on environmental impact and considerations with AI.

AI images

AI Image Editor Drift Test: 40 Edits, 4 Models

This article shares the results of what happens when you make 40 edits in a row with 4 different image models. Nano Banana 2.1 was the most accurate at the edits, but the background had visibly drifted from the original by edit 10. Ideogram and Flux had less background drift, but they weren’t as accurate with the edits.

On any model, branch instead of chaining. Make each edit from the original, or from a version you already approved, instead of from the latest result. A model that redraws the frame adds drift every round, so fewer rounds means less drift.

—Bilal Azhar

Working with clients

Let Your Clients Make Terrible Decisions

This is a topic that comes up regularly in conversations with other consultants and freelancers. What do you do when a client ignores all your advice and decides to do something that you know goes against best practices? The advice here is for freelancers in general, but it applies nicely to L&D consulting. Keep a paper trail of your recommendations, detach professionally, reclaim your time and energy, and be ready to send a proposal for the rescue mission later.

Your job as an expert consultant is to give them the best possible information, highlight the risks, and present strategic alternatives. Once you have done that clearly and professionally, you have fulfilled 100% of your contractual obligation.

If they choose to ignore your map and drive straight into a ditch anyway? That is their choice to make. You are not responsible for saving people from their own decisions after you’ve handed them the life vest.

—Sara at Freelancing Females

Rise course examples

Really Good Rise Courses | Maestro Community

Maestro (the company that makes the Mighty Rise plugin) has collected some examples of high-quality Rise courses. These feature their plugin, so it’s more than just the default Rise functionality and design. However, it’s a good place to get some inspiration.

L&D job searching

Learning & Development Professionals Career Opportunities Board

This LinkedIn group collects and shares L&D opportunities and helps job seekers connect with recruiters.

---

Upcoming events

DevLearn

Register for DevLearn by October 2 for the Advanced Rate. Save an extra $100 with code CHRISTY.

Branch Smarter, Not Harder: Naming Conventions, Variables, and Bottlenecks in Storyline

• November 3, 2026 at 9:20 AM
• CE01: Storyline & Rise Pro Lab

Cohesive by Design: Generate Consistent AI Icons, Illustrations & Characters

• November 4, 2026 at 10:00 AM
• BYOD (Bring Your Own Device)

Buzz Discussion: Video & Media

• November 4, 2026 at 3:45 PM

The post Video and Animation, AI Environmental Impact: ID Links 10/27/26 appeared first on Experiencing Elearning.

See how you do at identifying AI images with these pairs. I’m starting with the oldest pairs, which are over a year old (and may be easier to identify than the newest images).

Pair 1: Sitting in a coffee shop

Pair 2: Fall leaves

Pair 3: Hands

Pair 4: Conversation

Does it matter if you use real or AI images in training?

To some extent, maybe it doesn’t matter if it’s AI or not. With my most recent pairing, multiple people shared thoughtful comments about whether they would use either of the images for training, regardless of whether they’re AI or stock photos. A realistic looking AI image that matches the context of your audience may be more effective for supporting learning than a less relevant or fake looking stock image.

Making AI images look more realistic isn’t about being deceptive; it’s about generating something that looks like our audience’s workplace and that isn’t distracting.

Ironically, as AI gets better at generating photorealistic images, I find myself leaning more into less detailed illustrations. Now that it’s easier to get illustrations in a cohesive style on basically any subject, sometimes simpler images are better. They convey the meaning without distracting learners with too much detail.

Check the answers

Don’t peek until you’re ready!

• Pair 1 (Coffee shop): A is real and B is AI.
• Pair 2 (Fall leaves): A is AI and B is real.
• Pair 3 (Hands): A is real and B is AI.
• Pair 4 (Conversation): They’re both AI. A was generated with Nano Banana; B was generated with Midjourney (and I couldn’t get hands I was happy with on the woman in red after several attempts).

How did you do? Overall, most people on LinkedIn got these questions incorrect. Sometimes they identified the AI image but thought both images were AI when one was real. The only pair the majority of people guessed right was my original pairing of hands for pair 3. I swapped out the original hand for one I generated with GPT Image 2.5 Sunburst, the newest high quality image generation model from ChatGPT. This is the first image model that accurately generated someone writing with their left hand. Every other model I’ve tested previously has failed to show left-handed writing.

And more importantly: even if you were able to identify the AI images, would images like these be good enough to use in your training? What do you think?

---

Upcoming events

DevLearn

Register for DevLearn by October 2 for the Advanced Rate. Save an extra $100 with code CHRISTY.

Branch Smarter, Not Harder: Naming Conventions, Variables, and Bottlenecks in Storyline

• November 3, 2026 at 9:20 AM
• CE01: Storyline & Rise Pro Lab

Cohesive by Design: Generate Consistent AI Icons, Illustrations & Characters

• November 4, 2026 at 10:00 AM
• BYOD (Bring Your Own Device)

Buzz Discussion: Video & Media

• November 4, 2026 at 3:45 PM

The post Real or AI? Can You Identify AI Images? appeared first on Experiencing Elearning.

I published my original version of this post almost a year ago, and so much has changed since then! This is my update as of October 2026, with notes focused on how AI image generation tools work for L&D tasks.

What to pay for: The quick answers

I currently pay for both Flora and Midjourney. Flora gives me a wide range of image models and tools, and Midjourney gives me more interesting artistic styles for my blog headers.

One tool

If you’re going to pay for one tool specifically for image creation, I recommend you subscribe to one of the all-in-one multimodal tools like Flora or Magnific. These give you access to multiple frontier models plus additional tools to make professional work easier.

Two tools

If you can pay for two tools, add Midjourney or another specialized tool depending on your needs (Recraft, Brushless, Napkin, etc.) Which specialized tool is best depends on your specific needs.

Cost comparison

This comparison includes the lowest paid plan, which is generally enough for L&D work.

All of these tools except for Midjourney have a free trial. I recommend that you pay for one month and try a tool with your own projects before committing to a longer subscription.

Flora: Connect multiple tools for better workflows

If you do regular professional work generating images and videos, an all-in-one canvas tool like Flora provides a much better interface and workflow than working with a chat interface. Flora connects multiple tools into a single workflow, so you can switch between Nano Banana, GPT Image, or whatever image model you need. Unlike other tools, their pricing for each generation is completely transparent; I can see how much each image or video I generate will cost in actual dollars, not just tokens or credits.

With a node-based tool, you can combine multiple reference images and prompts and then generate and iterate multiple outputs. It’s also easier in this type of tool to do multiple types of transformations such as editing images, removing backgrounds, or adding text as a separate layer. You can easily switch modalities at different points in your workflow, so you can generate images with one tool and then switch to generating audio or video with other tools all in the same canvas or project.

Rather than generating each image one at a time, you can batch generate multiple images at once or use techniques to save a series of actions to repeat later. I used a list of poses to generate sets of character variations from a single reference image, all in a single step. Features like batch nodes and repeatable techniques can speed up your workflow.

Flora is the tool I use most often for generating images and videos for training. If you’re only going to pay for one image generation tool, I recommend a tool that gives you access to all the major models and additional features like Flora or Magnific. These give you the flexibility to use different image models plus all of the other features for professional work at scale.

Midjourney: Most interesting and artistic, style consistency

Midjourney is the tool I use most often for my blog header images. It’s the image generation tool that I find most interesting; it’s easier to generate images that don’t look like typical AI or stock images. You can create images in consistent styles and colors from any starting image, and you can use and generate style codes to expand your options.

Both the overall aesthetic of Midjourney and the infinite range of styles possible are reasons I see many top AI image experts start with Midjourney, even if they then move to other tools in their workflow. I generally have the highest success with Midjourney for generating visual metaphors and symbolic images to support learning. Midjourney is the best at artistic images, which means it’s a great choice for illustrations.

Midjourney handles photorealistic characters reasonably well, including generating multiple poses and expressions from the same starting character image. It’s not as good at consistency as Nano Banana, ChatGPT, or some other tools. I find that skin textures become less realistic over subsequent generations though. Midjourney’s editing tools like inpainting are useful.

Midjourney is better at generating text than it was a year ago, but still not effective enough to rely on it. For my blog headers, I just add the text in Affinity Designer so I can maintain consistency.

Gemini (Nano Banana): Best free tool, character consistency

Of the free tools for AI image generation, Gemini is probably the best overall option. It’s capable of generating and editing images and creating multiple poses of photorealistic characters. It’s easy to make the edits by describing them in plain language. You may run into limits or experience slower generation on a free plan.

Gemini is the tool; Nano Banana is the general name for the image models in the Gemini LLM and available through other tools. Although I work within Flora for the interface, I use Nano Banana image models most often for photorealistic images.

Nano Banana is good at generating consistent icon sets; you can prompt for a grid of icons in a single prompt and get results with high accuracy. It’s not as good at illustration styles as Midjourney; the styles tend to look more similar and more recognizable as AI.

Basically, Gemini’s image generator does all the things that ChatGPT’s image generation can do, but it does them all a little better. It doesn’t look as obviously AI, and it tends to be more accurate in following prompts both for initial generation and edits. I find that it still looks like stock images if you don’t prompt well. While the text accuracy is good, the fonts sometimes look out of place, especially on edits.

ChatGPT: Improving, but still clearly AI

ChatGPT’s latest image models, GPT Image 2.5 Flare and Sunburst, were released in September 2026. These models are both significant improvements from the early models. They’re more accurate in following prompts and editing parts of an image without changing the rest. You can generate both photorealistic and illustrated images.

One of the challenges of ChatGPT for images is that it’s such a common tool that’s it’s very recognizable. Especially for images with text, the default style is instantly detectable as ChatGPT from the paint slash with white text on it on the “whiskers” emphasizing text. A lot of what people perceive as “AI slop” posters is generated quickly with ChatGPT.

It’s possible with better prompting to get better results that don’t look so distinctively like ChatGPT, but it’s more work. I still generally prefer Nano Banana to either of the GPT Image models, but it’s much closer now.

Magnific: Multiple tools in a combined workflow

Magnific (formerly Freepik) is similar to Flora. It offers you a choice of multiple tools on a visual canvas where you can easily see your work and switch tools as needed. They’re regularly adding new features for creative professionals such as a brand kit to save your colors and other brand details.

I tried both Magnific and Flora, and they’re both great options. Personally, I find Flora’s canvas to be a little smoother to move around; it’s easier to switch between panning on the canvas and adding nodes. But if you’re interested in this kind of tool, it’s probably worth trying each one for a month. Depending on your specific workflow or how you collaborate with others, Magnific might be the better option.

Recraft: Brand consistency, text accuracy, icons, and vector images

Recraft is a tool that should definitely be on your list to try if you need consistency in branding or need to generate vector images. Many image generators create images that look like vectors but aren’t actually SVGs; Recraft generates true vector images so it’s easier to scale and edit them elsewhere. They have a tool for generating icons that lets you generate up to six icons at a time, all with a consistent style and color palette. Their text accuracy is also much higher than other tools.

I don’t think their photorealistic images are as good as Nano Banana or Midjourney, but it’s a solid choice for illustrated images. You can use the built-in styles or create your own. It even has options to share styles within teams, which makes it a better option for organizations with multiple designers.

Brushless: Vector images and icons with consistent styles and colors

Brushless is another great option for vector images and icons. They have several illustration styles that are ready to go. I think their built-in flat icon style works well for training purposes, especially when you need to generate specialized icons for presentations or highlighting important concepts. The line art style is good for simple illustrations where you don’t want to distract learners with overly complex images (and AI images can be overly complex and distracting). You can set your color palette to match your branding and use reference images to create your own custom styles.

Napkin: Images and visual organizers to explain concepts

Napkin works differently than a lot of the other AI image generation tools out there. Instead of prompting for images, you paste your text such as bullet points or a paragraph of explanation. Napkin interprets that text to make an image to explain it. The result is something like PowerPoint’s SmartArt but better. You get icons for main points and diagrams to show connections and structure illustrating your concepts.

Ideogram: Image editing

Ideogram was one of the first AI tools to figure out how to render text accurately. Ideogram generates some interesting illustrations, so sometimes I use it to compare how it does with the same prompt as other tools. Ideogram has image editing tools that can be useful to refine and extend images, even from other sources. Ideogram’s most recent update focuses on editing just a portion of an image without regenerating the entire image. Most AI tools drift over the course of successive edits because they regenerate everything. If you need a background to stay the same while you edit a central object or character, Ideogram is worth trying. Their lowest paid plan also include unlimited slow credits, which means you can keep generating without limits (just not very quickly).

Hands-On AI Image Workshop

Want to learn how to generate higher quality AI images—visuals that are good enough to actually use in a course? Part of the trick with AI image generation is that even if you start with a good prompt, you still may need to iterate a few times to refine it. Even though I’ve presented on AI images multiple times, it’s impossible for me to give participants enough time to really practice that iteration process in a one-hour webinar.

Registration is now closed for the workshop, but you can sign up for the waitlist to be notified when this workshop is offered again.

Learn more about the AI Image Workshop

These are my opinions

None of this is sponsored; it’s all based on my own experiences. In a few cases, I have gotten some extra free credits with specific tools in order to test them or in exchange for providing feedback on the tools. I have paid subscriptions to Flora and Midjourney, and I get access to a number of other image generation tools through my subscription to Poe.

Your experiences may be different. Like all AI tools, the image generators are sometimes unpredictable in what they can do effectively and where they fail. You may find that different tools work better for your specific needs.

For all of these tools, check the terms of use. Most explicitly allow commercial use of images on paid plans, but not necessarily on free plans. Some tools will make your generated images public, especially on a free plan. Some will train their AI on your images, including uploaded reference images.

More tools?

Obviously, I didn’t cover every AI image generation tool currently available in this post; this is based on what I’ve personally tried.

• Same Prompt, Different AI Image Models
• Consistent Character Images Step by Step using Flora and Nano Banana
• Updating Scenario Character Images with Flora
• Speed Up Character Image Creation with Flora
• AI Images with Consistent Style in Midjourney
• Testing Freepik Spaces for Character Image Generation (It was Freepik when I tested it, but Magnific now.)
• Nano Banana (Gemini) Character Image Experiments
• Frustrations with ChatGPT Image Generation
• How to Generate Icon Sets with AI using Midjourney, Nano Banana, ChatGPT, and Brushless

Originally published 11/4/25. Updated 4/28/26, 10/12/26.

---

Upcoming events

DevLearn

Register for DevLearn by October 2 for the Advanced Rate. Save an extra $100 with code CHRISTY.

Branch Smarter, Not Harder: Naming Conventions, Variables, and Bottlenecks in Storyline

• November 3, 2026 at 9:20 AM
• CE01: Storyline & Rise Pro Lab

Cohesive by Design: Generate Consistent AI Icons, Illustrations & Characters

• November 4, 2026 at 10:00 AM
• BYOD (Bring Your Own Device)

Buzz Discussion: Video & Media

• November 4, 2026 at 3:45 PM

The post Comparing AI Image Generation Tools appeared first on Experiencing Elearning.

Step 1: Reference image

The first step to generating a set of consistent character images is generating a reference image. Sometimes I initially generate a character on a blank background for my reference image, but this time I just started with an image of the character in a scene. For a larger set of images, a full body reference image or a character sheet from multiple angles is helpful. I knew this was going to just be a small set though; I was experimenting with camera angles for a blog post.

I used my standard image prompt framework to specify the image style, subject, action, setting, and details. All of the images in this post were generated in Flora with the Nano Banana 2 model.

Editorial photo, Puerto Rican man in his 20s with a beard trimmed close, wearing a dark orange shirt and blue jeans, black glasses, sitting on a bench at a park looking at his phone, low camera angle

Results: As you can see, this image didn’t quite work. What’s going on with that wooden half bench back or fence in front of him? I like the character, but the physics doesn’t make sense.

Using the initial image as a reference image by attaching it to the prompt, I asked for a new image.

Pan out to a wide angle establishing shot of the park with this man sitting in the bench viewed from farther away

Results: Just panning out and changing the scene got rid of the weird wooden half bench in front of it. It changed several other details in the image as well: he’s no longer smiling, there’s now a section of green grass where people are sitting instead of trees. It’s still the same person with the same outfit though; the character consistency is good.

I wanted to see what this would look like with more grass instead of his bench being on a paved area.

Extend the grass to the bottom of the image so the grass is under the bench. Keep everything else the same.

Result: This gave me a weird patch of grass underneath. I didn’t like this iteration, so I discarded it and decided to stick with the previous one.

Step 2: Change the action

With my initial character reference ready, I could start generating additional images of this character. I used my very first image as the character reference here since it was a closer image. I changed the action and the camera angle. Because I attached the reference image to the prompt, all I needed to say was “this man,” and Nano Banana maintains the details.

Create a close shot, side angle, profile view of this man talking and explaining something, head and shoulders only. 

Result: I liked how the character turned out here. His pose looks fairly natural, and his skin texture looks fairly realistic even in this close up. Because I used that very first image as my reference, you can still see the weird wooden half bench in the background, but at this angle it just looks like he’s standing near a bench.

Step 3: Change the camera angle

Next, I wanted an image where I changed the camera angle. I wanted to show him talking to another character, but with the camera behind him. I used the image of him talking as my reference image. This means I’m only changing one thing: the camera angle.

Create an over the shoulder image showing the back of this man's head while he's talking. 

Results: The camera angle change worked on the first try. I’m happy with the consistency of the character here.

Step 4: Add another character

Now, I needed another character in the scene so it’s clear who he’s talking to.

Zoom out to show that the man is talking to a Chinese woman who is listening with a neutral expression. She is wearing a blue shirt.

Results: Success in one try! The AI panned out the camera to give more space for the second character while maintaining the first character.

I might have been able to do steps 3 and 4 together, but I find that I often get better results when I limit the edits to one major change at a time.

Step 5: Change the setting

I also wanted an image of this character walking. For my first try, I kept the scene in the park.

Create a tracking side angle image of this character walking on a paved trail in a park, afternoon sunlight

Results: If I only needed a single image, this one might have worked OK. But when you see it side-by-side with the image of him sitting on the bench, you realize that the background is the same. I didn’t tell the AI to choose a different part of the park, so it created an image by removing the bench and keeping the character in the same location.

Rather than trying to get an image in this park, I decided to change the scene entirely and put him walking on a street.

Change the background to be a quaint main street of a small town in Florida

Results: The character consistency is good, but the setting has multiple problems. What’s with the weird vintage cars, especially the brown one in the back? And the name of the restaurant is the…”Florida Cracker Cafe”? Um, that’s certainly an interesting choice that the AI made. This is a great example of why you need to check your AI images, including the background details. I decided to leave the greenery in front of the restaurant, but I often iterate images to tell the AI to remove the proliferation of potted plants.

In my prompt for changing the text, I put the exact words I wanted on the sign in quotation marks. I made two edits at once here. If that had failed, I would have gone back and split up the changes across two edits.

Change the name of the "Florida Cracker Cafe" to "Sunny Crepes and Coffee". Remove the cars on the street.

Result: This one looks much better. I could still do more with the other signs in the background. It depends how you’re using an image how much you need to zoom in and inspect it. This image series was for a blog post banner, so I knew I wasn’t going to use them in a large format. In a small size, the background text is probably OK as is. If this was a hero image in a course, I would spend more time polishing those details.

Step 6: Review all images

After finishing the entire image set, I looked at them all together one more time to make sure I didn’t need to make any further adjustments. Here’s what the whole set looks like in Flora. I generated 9 images and kept 4. In the screenshot below, you can see my workflow, moving generally left to right. Each connecting line shows where I used a reference image.

Bonus images

I also generated two more images later in response to someone’s question about images of people opening doors and windows. Both of these images have some issues, but opening windows was particularly challenging to get right. Because the AI image models don’t generate images based on actual physics, they get these sorts of details wrong regularly.

Edit and iterate to improve your AI images

A lot of AI images look really obviously AI. Partly, that’s because people generate something and just accept the first result. Better initial prompting and providing design guidelines can improve that first result, but you still need to edit and iterate. I hope this worked example of my step-by-step process of creating consistent character images helps you learn to edit and iterate your own images more effectively.

AI Image Workshop October 14

If you want more practice and guidance on this process of editing and iterating, sign up for my AI image workshop, From First Try to Finished. You’ll get six rounds of hands-on practice, much of which includes this kind of image refinement like I showed in this post. We’ll debrief each round of practice to talk about what worked and what didn’t, and I’ll provide feedback and coaching on how to improve your results. You’ll get 3 hours of specific tips I’ve learned through doing it myself plus time to immediately apply what you learn.

Registration closes Tuesday, October 13.

The post Consistent Character Images Step by Step appeared first on Experiencing Elearning.

Accessible color

Accessible Color Choices Are About So Much More Than Just Contrast – Sheri Byrne-Haber’s Blog

This article is a deep dive into accessible color choices for potential issues such as high contrast repeating stripes causing migraines or saturated red triggering seizures. This includes a list of color issues not regulated by standards but that may still cause problems (such as cultural differences in colors). h/t Kayleen Holt

APCA Contrast Calculator

This contrast checker uses a different calculation than the standard WCAG color contrast checker. It recognizes a difference between light text on a dark background and dark text on a light background. Sometimes color pairs that meet the WCAG can still be hard or tiring to read, and this calculator may be especially useful for those pairs.

A Deep Dive into Contrast Themes | Digital Accessibility Services

Contrast Themes are an accessibility feature in Windows that overrides the default colors used to provide custom choices for better accessibility. This is beyond doing a light and dark theme, but a way for the user to adapt the app or website in a way that works for them. This article gives specific tips for better compatibility with Contrast Themes. For example, if the focus state would change background color but not outline, using a transparent outline instead of outline: none lets the Contrast Theme provide an outline in a user-selected color.

AI images

Welcome to Magic: The Editor – by Humble Nations

Midjourney’s new editor is now available. You can prompt for edits with natural language, more like prompting edits in Nano Banana. This article includes a bunch of examples of editing prompts for changing the style of images, plus a few others like changing camera angle, character pose, and prompting layouts from sketches.

GPT-Image2 Prompt Gallery

A collection of hundreds of examples of images and prompts created with ChatGPT. Filter the list by category and style. The examples are sometimes uneven in quality; some of these have that very obviously ChatGPT look to them. But there are some creative ideas too, and it’s helpful to see examples of prompts to see how to describe different visual styles and elements.

Why A.I. Struggles with Negative Words | Otherwords

If you’ve ever tried to generate an image with AI and asked for “no plants” in the office, only to get a bunch of plants in the image, you’ve seen how much trouble AI has with negative language. This video explains why negative words are hard for AI to understand.

Real or AI? – Test Your AI Image Detection Skills

Can you identify which images are real and which ones are AI? I did terribly on this quiz the first time I tried: less than 50% right.

AI video

SLICE: how we made a 2-minute AI short film with Seedance 2.5

This is a great example of a “show your work” post on what it takes to generate high-quality AI videos. The prompts to create the characters and scene and to generate the video clips are included. The amount of detail needed to prompt in Seedance is significant; the prompt for the first 30-second clip is 700 words long. Check the notes at the end about what failed and how he worked around the problems too.

Remove Text from Video Online – AI Video Text Remover

I’ve never personally had a need to do this, but I can see how this tool might be useful. If you have a video with burned in “open” captions, this AI video editing tool says it can remove the text. The pricing options include buying a credit pack, so if you only have an occasional need for this you can buy credits instead of adding yet another monthly subscription.

Working with AI

The AI Workflow Redesign Method: How the Top 6% of Users Get Real Returns from AI

Dr. Philippa Hardman explains her process for analyzing a workflow to determine how to redesign it to get more value out of AI.

Psychological Stamina and other #NerdStuff – Usable Learning

Julie Dirksen explains the concept of psychological stamina. Even if you have the knowledge and skills to do a certain task, how long can you keep doing it? If the behavior requires vigilance over time, like watching for hazards on the road while driving, how long can you maintain that level of mental focus?

Diligence for reviewing outputs is a key element for successful use of AI (as I may have mentioned a few times before).

—Julie Dirksen

AI environmental impact

AI vs. Streaming: Energy, Water and Hidden Costs

Josh Penzell created a calculator to help people compare the energy, water, and other environmental costs of using AI in comparison to streaming. One insight I had from this tool is that AI uses a lot more direct on-site water. That’s probably where some of the really high numbers of AI water use come from. But when you look at the total operational water including the electricity use, then streaming uses more water than AI. All of the source data is provided so you can dig into it yourself.

Tools for instructional designers

The ID Toolbox: free tools for instructional designers

A collection of free utility tools for instructional designers. I will definitely use the Caption Line Splitter, which rewraps captions based on the number of characters and lines. The collection includes other utilities for accessibility, QA, calculating development time, identifying which text will overflow when localized, and more.

Hands-On AI Image Workshop

Want to learn how to generate higher quality AI images—visuals that are good enough to actually use in a course? Part of the trick with AI image generation is that even if you start with a good prompt, you still may need to iterate a few times to refine it. Even though I’ve presented on AI images multiple times, it’s impossible for me to give participants enough time to really practice that iteration process in a one-hour webinar.

Registration is now closed for the workshop, but you can sign up for the waitlist to be notified when this workshop is offered again.

Learn more about the AI Image Workshop

The post Accessible Color, AI Images and Video: ID Links 9/29/26 appeared first on Experiencing Elearning.

Use reference images

My first lesson learned was to use reference images. I already knew that, to some extent, because I use reference images for character images. I used those same character reference images for all of the videos you’ll see in this post.

However, I didn’t use a reference image for the setting. While I did have images I could have used for the first frame, I’m still not completely happy with what I’d gotten previously. I decided to see how the different tools did at setting the scene without a reference image. That was OK for this experiment, but I think for further development I’d want to have at least one reference image for the setting, if not a storyboard with multiple images.

Instead of a reference image, I included a description of the room in my prompt. I include a description of art on the walls because otherwise the AI tools often fill in other details. I often get unrelated text on a whiteboard or charts on a screen, and I didn’t want those here. (Again, if I’d used a reference image for the room, that would have fixed the problem.)

The dialogue comes from the opening passage of my project management scenario.

Setting: Three people @Jake @Samantha @Alex meeting in an office conference room, all sitting around a table. Laptops and coffee mugs on the table. Light gray walls, overhead lights, abstract corporate art on the walls.

@Jake : "Alright team, we're a week behind schedule on the Healthtrack project, and the client still hasn't provided the data integration requirements we requested three weeks ago."

@Samantha: "Without those requirements, we can't properly design the system architecture. We risk making assumptions that could lead to costly rework down the line."

@Jake "We can't afford to keep falling behind. We should make our best educated guesses based on our experience and move forward."

@Alex "I'm with Samantha on this. If we don't get the integration requirements right, it could severely impact the user experience and data flow."

This first example video was generated in Google Omni 1.1 Flash. I’ve had decent luck with this model for generating b roll videos.

• Model: Google Omni 1.1 Flash
• Video length: 10 seconds
• Cost: $1.20
• Cost/sec: $0.12

All costs noted are my actual costs as calculated in Flora; I did math to determine the cost per second.

Give enough time to include the dialogue

I immediately realized one of my errors: that dialogue was too long for the generation time. The video is 10 seconds, but I had more dialogue than could fit. The AI trimmed my script to fit it in rather than using it verbatim. That would be fine for some uses, but that’s not what I wanted for this scenario.

For my next attempt, I trimmed it to just the first line of the script. I hoped that would give it enough time to keep the dialogue accurate. And…well, you can see for yourself. It was closer, but it still wasn’t quite right. I also adjusted the description of the room; I decided I didn’t like the abstract art.

Setting: Three people @Jake @Samantha @Alex meeting in an office conference room, all sitting around a table. Laptops and coffee mugs on the table. Light gray walls, gray chairs, light wood table, a clean whiteboard on the wall. Overhead florescent light.

@Jake : "Alright team, we're a week behind schedule on the Health track project, and the client still hasn't provided the data integration requirements we requested three weeks ago."

• Model: Google Omni 1.1 Flash
• Video length: 6 seconds
• Cost: $0.72
• Cost/sec: $0.12

Watch for character positions changing

One of the other problems in this video is that Alex shifts position. At the beginning, he’s sitting next to Samantha. Partway through, he magically transports to sit next to Jake. This is another problem I probably could have solved with reference images and a storyboard. I’ve even seen a demo where someone blocked the character positions in Blender for a rough 3D model and then used that to keep each character in the same place around a table. Keeping the characters and objects consistent from scene to scene is a challenge unless you’re directing it.

Some models look more AI than others

I wanted to test several models to be able to compare the results. I used the same prompt again with a different model, MiniMax. These results were even worse; this looked and sounded really obviously AI. The character and dialogue accuracy was lower. Check out how Jake’s background suddenly switches to the studio photo background rather than the conference room background. MiniMax was fast and reasonably cheap, but it’s not the right model for this task.

• Model: MiniMax H3 Max
• Video length: 8 seconds
• Cost: $0.768
• Cost/sec: $0.096

Video generation costs can add up quickly

Next, I tested Seedance. This is a higher quality model, but it’s also slower. Each generation took several minutes, where the previous videos tool under a minute to generate. I have had some good results with Seedance, but it’s also about four times more expensive than the Google or Minimax models. With Seedance, I generated an initial clip of 8 seconds. Then, I used the Extend video function in Flora to generate a second clip for the next clip with Samantha’s response. I stitched the two clips together in Flora.

• Model: Seedance 2.5
• Video length: 21 seconds
• Cost: $3.914 + $6.337 = $10.251
• Cost/sec: $0.488

You can see the Seedance clips combined below. The transition is weird; I should have prompted for the camera angle to change to point at Samantha instead of starting zoomed out and pushing in again. But because Seedance is so expensive, I was reluctant to generate another clip. If this was a real project and not for my blog, I would have had to pay the $6 for another generation.

While Seedance is higher quality (and I know I could improve my results with more detailed prompts and direction), the cost is significant. A 2-minute video would cost about $58 in generation. Knowing that some clips wouldn’t work on the first try and would need to be regenerated, a more realistic estimate for a 2-minute video might be about $75-$80. Maybe it’s because I’m used to how cheap image generation is (images cost 5-7 cents on Flora), but I was surprised at how fast the costs added up.

Voice quality and consistency concerns

AI voice overs have dramatically improved in quality in the past few years. However, in these AI generated videos, you probably noticed some issues. For example, in the Seedance clip above, Samantha has weird, overly dramatic pauses in her line. I’ve also heard issues in keeping the voice consistent for characters across multiple clips.

In order to generate enough videos for a full interactive video scenario or a longer conversation, I think I’d need to generate the audio separately in ElevenLabs or WellSaidLabs. That would give me higher quality and more control over the generation. Plus, that would fix the consistency issue; as long as you use the same voice for each character, you maintain the same sound throughout.

However, bringing in audio also means dealing with lip syncing in the video tool, and that creates a different set of challenges. How accurate is the lip syncing? How many generations does it take to get it right? After you generate clips, how much manual editing is needed?

Reflections on AI videos for training

This is a sample project for my blog and portfolio, so I’m not persisting through challenges the same way I would if this was a real client project. Right now, I feel like the AI video tools like this are good for generating short b roll clips. I’m already using AI videos for that purpose in actual explainer videos now.

For more complicated videos, it’s going to take a lot of time, iteration, and cost. There’s no magic where you drop a script in, press a button, and get a fully formed video with multiple characters, dialogue, camera angles, and consistent background out. It’s possible to generate higher quality videos with AI (check out how the 2-minute animated video Slice was created), but it takes quite a bit of work. I’ve seen some 700+ word long prompts for Seedance; that’s a lot of time directing the AI that I didn’t spend for these samples.

When my clients have asked me about AI video generation over the past two years, I’ve been telling them to wait another 6 to 12 months and then check in again. At this point, I think it can be done, but we have to be realistic about what it takes. All of these models are cheaper than hiring actors for a half day shoot, but it’s expensive and difficult enough that I’d need to scope it realistically.

Will you know that all of these are AI-generated videos? Yes. But getting the quality good enough to be useful for training scenarios doesn’t necessarily mean it has to be indistinguishable from using human actors. In fact, I think sometimes we’re better off using illustrated or animated styles that don’t pretend to be real people. You can still get the value of a conversational example or interactive video with something illustrated.

AI video tool comparison

Here’s my summary of the three AI video tools I tested.

Hands-On AI Image Workshop

Want to learn how to generate higher quality AI images—visuals that are good enough to actually use in a course? Part of the trick with AI image generation is that even if you start with a good prompt, you still may need to iterate a few times to refine it. Even though I’ve presented on AI images multiple times, it’s impossible for me to give participants enough time to really practice that iteration process in a one-hour webinar.

Registration is now closed for the workshop, but you can sign up for the waitlist to be notified when this workshop is offered again.

Learn more about the AI Image Workshop

The post Lessons Learned while Generating AI Videos appeared first on Experiencing Elearning.

Typical compliance training: Push

Let’s say you’re creating training for managers on providing reasonable accommodations for disabilities. A typical elearning compliance course on reasonable accommodations would start with a history of the ADA, ADAAA, Civil Rights Act, etc. (Readers outside the US, substitute your own laws about disabilities.) You might compare how the laws have evolved over time, define the terminology, and list some common reasonable accommodations you could provide.

All of that is “pushing” content to learners. The learners are treated as passive receptacles for content. You as the instructional designer control what content is delivered, in what order, and how quickly. The learners don’t make any decisions, and they don’t have any reason to dig deeper into the content.

Alternative compliance training: Pull

You could flip the whole course around though. Instead of pushing content to learners, you let them “pull” what they need. The trick is that you have to give them a reason to need the information.

In the example below, Rosa asks her manager, Cindy, for time off after a surgery. This is a plausible situation managers might find themselves in where the right answer might not be obvious. The managers have to make a decision: to grant the time off or not.

In this scenario, you could just offer the two choices: grant the time off or not. However, I’m deliberately putting them in this situation to give them a reason to look up the policy. Therefore, I have two more choices below: “Ask HR” and “Review Policy.” If learners choose “Ask HR,” they’ll see a simulated conversation with an HR specialist explaining the policy and providing advice. Either way, they’re “pulling” the policy information because they have an immediate need for it.

“It’s easiest to pay attention to content that you can use right away. Use strategies like test-then-tell, scenarios or problem-based learning to create an immediate use for the learning content.”

Julie Dirksen, The Science of Attention and Engagement for Learning

“Pull” options

You might choose to only have one option for explaining the policy rather than two like the above example. Either just reading it or asking another character for advice might be enough. If reading the policy is the only option, it’s helpful to include a plain language explanation. This is especially true if the original policy is written in very formal or legal language. You can provide the original and the explanation side-by-side, or you might just provide the explanation.

If your users will have a job aid when they actually do the task, use that job aid in the course so learners practice with that tool. If they won’t have a job aid in the real workplace, what would they do instead? Would they consult with HR or ask a manager? Figure that out, then build that into the scenario. As much as possible, make the way learners access policy information in your course similar to how they will access it on the job.

Knowing when policies apply

Another advantage to the pull approach is that it helps learners see when policies apply to different situations. This is especially helpful if your audience needs to be able to distinguish between similar situations. Maybe under most circumstances, “guideline A” applies, but there’s an exception where “guideline B” applies instead.

“Learners are motivated when they can see the relevance of what they are learning. If they learn new information in the context of real world problems, if they are shown a demonstration of skills they are expected to acquire (show), if they have the opportunity to apply these skills in the solution of additional real-world problems or the performance of additional real-world tasks (do), then they are much more likely to acquire, retain and be able to use the required skills.”

M. David Merrill, Finding e3 (effective, efficient and engaging) instruction

Another example: Cathy Moore’s Connect with Haji Kamal

Although Cathy Moore’s Connect with Haji Kamal isn’t compliance training, it also demonstrates the technique of having characters ask for advice. Learners need to “pull” information and actively seek it out rather than passively waiting for the course to “push” it to them. It’s the same mechanism, just a different kind of training.

Cathy’s original version is no longer available since it was built in flash, but Stephanie Harnett built The Return of Haji Kamal in Storyline with the original assets (and Cathy’s permission).

Try one-question mini-scenarios

Even with one-question mini-scenarios instead of branching scenarios, you can use this technique to motivate your learners to look up the policies themselves. Give learners a reason to seek out and immediately apply the information. That will make them more likely to remember those policies when similar situations happen on the job.

Image credits

• Characters from eLearning Art
• Document by useiconic.com from the Noun Project
• Conference room from Storyblocks

This compliance course example is also used in What’s Your Character’s Goal?

Originally published with the title “Motivating Learners to Look Up Compliance Policies Themselves” on 3/22/2016. Updated and republished with a new title 7/21/2020. Last updated 9/10/2026.

The post Create Motivating Compliance Training appeared first on Experiencing Elearning.

You’ll learn strategies for writing more effective image prompts to minimize bias and reflect more inclusive representation in your learning. You’ll get tips to more authentically represent your training audience by using AI images. While AI tools are improving all the time, they still have limitations that you need to be aware of. We’ll discuss some of the ongoing challenges with inclusive representation in AI image generation and options to minimize those issues. You’ll leave with a collection of sample image prompts, a prompt framework for better results, plus tactics to make your AI-generated images more inclusive.

Watch the recording

If the video isn’t embedded below, you can watch the presentation recording on YouTube.

Download the slides

Most of the time, I don’t share my slides for presentations. My slides aren’t all that useful on their own, so I build separate handouts with more comprehensive information. However, in this case, where the focus is often on both the inclusive AI images and the prompts to generate them, having the images of the slides is helpful.

AI still can’t show left-handed writing

One of the observations I shared in my session is that AI image generators still can’t show someone writing with their left hand. That’s partly because the majority of images of people depict right-handedness. That becomes the default prediction for the AI when showing writing, and I haven’t found any AI image model that can override that default.

For example, here’s Nano Banana’s result for this prompt:

Editorial photo, a left-handed woman writing in a notebook with her left hand in a charming coffee shop, focus on the hand, notebook, and table. An iced latte is on the table next to the notebook.

Is left-handedness the most pressing issue related to inclusive AI images? Of course not! But it’s an example of how AI predictions about the most likely results can amplify stereotypes in images. The same types of predictions happen with gender, race, ethnicity, nationality, hair, skin color, disabilities, assistive devices, body type, and more.

Generating images of a range of body types

AI images tend toward slender body types. Adding descriptive phrases like “average weight” or “slightly overweight” can help counter that. 

These four images all used the same prompt except changing the body type as shown.

Style: Candid iPhone photo
Subject: Greek woman, mid 30s, black wavy hair, olive skin
Body type: [Very thin, athletic; Average weight; Slightly overweight; Obese, heavyset]
Clothes: Blue and white floral shirt, tan slacks
Action: Laughing at a joke, pointing
Setting: On the sidewalk, outside a cafe
Lighting: Golden hour
No signs or text

More on inclusive AI images

• Disability Representation in AI Images
• Nonbinary Characters in Scenarios
• Considerations for Using AI Images Ethically
• Minimizing Bias in AI-Generated Images (My AIDC presentation last year)

---

Upcoming events

DevLearn

Register for DevLearn by October 2 for the Advanced Rate. Save an extra $100 with code CHRISTY.

Branch Smarter, Not Harder: Naming Conventions, Variables, and Bottlenecks in Storyline

• November 3, 2026 at 9:20 AM
• CE01: Storyline & Rise Pro Lab

Cohesive by Design: Generate Consistent AI Icons, Illustrations & Characters

• November 4, 2026 at 10:00 AM
• BYOD (Bring Your Own Device)

Buzz Discussion: Video & Media

• November 4, 2026 at 3:45 PM

The post Inclusive AI Images for Learning appeared first on Experiencing Elearning.