We did it!!!
Mel Milloway shares an example of an interactive avatar for interview practice. Mel has been working out loud throughout the process of building this project to explain her work and the challenges she ran into throughout
This Image Wasn’t a Stock Photo – and It Changed the Way I Build Training
Michelle Bonkosky shares her process for using ChatGPT for generating a unique image for a training workbook. I appreciate how she shows the process of iterating and refining her prompts; that's a key point. She also includes some sample prompts for images for training assets.
AI Co. Anthropic Nabs Partial Fair Use Win in Copyright Case
The headlines about this case will miss a lot of the nuances; it's not a complete win for Anthropic, but it is an important one. The ruling found that training AI on legally obtained copyrighted books is fair use because it's "quintessentially transformative." That doesn't mean that training on pirated books is fair use, and nothing in this ruling explicitly addresses content publicly available online. The output of AI is also an unresolved question; I predict we'll have some rulings that generating text or images that too closely matches existing copyrighted works is not protected. AI tools (especially image generation tools) need guardrails to prevent the generation of copyrighted content.
Gemini Image Editing: A Guide With 10 Practical Examples
This article explains how to use Gemini for image editing with examples, showing what works and what doesn't work. Modifying facial expressions and body poses worked fairly well in this author's testing, making this one option for generating sets of character images.
Bias in AI: Examples and 6 Ways to Fix it in 2025
Examples of bias in AI in image generation, recruiting tools, voice recognition, and other areas. The solutions here focus primarily on adjustments of the AI systems and debiasing strategy rather than on the level of individual prompts to improve representation. If you're looking at your overall strategy for AI, how you address bias has to be part of the plan.
Why AI Video Avatars are NOT the Next Big Thing in L&D
Heidi Kirby digs into the research about AI video avatars (excluding the vendor research). The support really isn't there. I've anecdotally seen lots of complaints about how they sit in the uncanny valley. But even as the video avatars get more realistic, is a talking head video really the best instructional method? Of course not! There wasn't a lot of buzz about talking head videos before AI. Why is there so much buzz now? (Interactive video avatars for scenarios are a separate question and not addressed by this article.)
Despite their increasing use, there's limited evidence that AI-generated avatars significantly improve learning outcomes.
Lemon Slice
Create talking characters and avatars with this AI video tool.
Creating with Gen-4 Image References
Runway Gen-4 offers a more controlled workflow for generating images with consistent characters and scenes. Add either a single reference image of a character or multiple reference images to combine multiple characters in a scene or specify the setting or objects. This is a more complex workflow than just chatting with ChatGPT, but it gives you more precision and more consistent results. This is Runway's documentation on using image references.
Supporting Learning with AI-Generated Images: A Research-Backed Guide - MIT Sloan Teaching & Learning Technologies
Suggestions and examples for using AI-generated images in meaningful ways to support learning, without adding confusing or distracting images. Consider cognitive load and the purpose of your images.
A study by Sung and Mayer (2012) suggests that any graphic in a learning experience will fall into one of these three categories:
Instructive images: These visuals directly support learning and facilitate essential cognitive processing of core concepts. For example, a diagram illustrating Porter’s Five Forces can help students better understand this business strategy framework.
Decorative images: These graphics enhance aesthetics but don’t influence learning. For example, an image of a business handshake can be visually appealing but won’t support or obstruct students’ understanding of negotiation strategies.
Distracting images: Sung and Mayer call this category “seductive” images. While these visuals may relate to the topic, they impede learning because they require extraneous cognitive processing. As an example, consider a complex organizational chart of a full corporation in a lesson on team leadership. The image connects broadly to the lesson but also highlights a lot of irrelevant details, distracting students from the key concepts.
The recent history of AI in 32 otters
Ethan Mollick shows the progression of AI image and video generation with iterations of a prompt about otters using wifi on a plane. He also explains the difference between diffusion and multimodal image generation models (Midjourney vs ChatGPT). These tools get such different results because the underlying technology and approach is different.
While LLMs generate text one word at a time, always moving forward, diffusion models start with random static and transform the entire image simultaneously through dozens of steps. It is like the difference between writing a story sentence by sentence versus starting with a marble block and gradually sculpting it into a statue, every part of the image is being refined at once, not built up sequentially.
But what makes diffusion models interesting is not their increasing ability to make photorealistic images, but rather the fact that they can create images in various styles.
Unlike diffusion models that transform noise into images, multimodal generation lets Large Language Models directly create images by adding tiny patches of color one after another, just as they add words one after another. This gives AIs deep control over the images it creates.
Rime | Voice AI
The advances in AI voices continue to impress me. Rime is aimed more for organizations using AI voices for customer service or live conversations, so it might be useful for voice chat in training applications. There's a free plan available to test it out.
AI image generators tend to exaggerate stereotypes
The examples in this article are all from older images, but the problems of bias in AI image generators remain. Unless you are explicitly prompting to avoid stereotypes, AI image generators reflect the bias of the images they trained on. Even if you do prompt to avoid stereotypes, it can still be a problem.
Demystifying LoRAs: What are they and how are they used in Stable Diffusion?
Quick introduction to LoRAs and how they're used for fine tuning image generation in different styles for Stable Diffusion (and other image generators)
How to achieve character consistency
How to video from Flora about how to create images with character consistency across different scenes. This is a more time consuming and technical process involving training a LoRA (low-rank adaptation) on an initial set of images for a character. This probably works best with real people, but there may be ways to adapt this workflow for elearning with generated characters. This is more effort than I would do for most projects, but might be worth exploring if I need something higher end for a specific project.
Podcast Transcript AI - Transcribe Any Podcast For Free!
Generate a transcript of any podcast on Spotify or Apple Podcasts. Search for the name of the podcast, pick an episode, and get a transcript emailed to you. I wanted to get a transcript of one of my podcast interviews, and this was a quick way to generate one.
The Human-AI Task Scale | Josh Cavalier
Josh Cavalier shared his Human-AI Task Scale showing a continuum of how humans and AI can work together. This makes so much more sense to me than a simple binary of "do it all yourself" or "AI does everything." This reflects much more range and shows opportunities for different kinds of tasks and work.
AI Art Generator: Free AI Image Generator & Editor | OpenArt
Generate consistent character images in multiple scenes starting from a single image. You can also use image to video tools.This integrates with other tools and gives you the option to train your own AI model with your style for illustrations. The free plan is limited, but there are paid plans at different levels.
Say What You See - Google Arts & Culture
Learn how to write better prompts for images with this tool. Describe what you see in an image and see how close your generated image is to the original. This tool uses AI to analyze your results and how accurate you were.
Midjourney’s Image Editor Update: NEW Layers, Smart Selection, and More
Directions on how to use the layers and Smart Selection features in Midjourney for more editing control. This gives you more options to keep the parts of an image that you want but to regenerate or edit variations. You can also assemble scenes with layers.
FLORA
Combine AI text, image, and video generation tools together for more complex workflows. I haven't tested this tool yet, but it might be worth using the free plan to experiment and see what's possible.
FLORA + Gemini - Multi-Character Scenes
Rory Flynn (@Ror_Fly) on X shares a basic workflow for generating character images in Midjourney and using Flora and Gemini to create scenes with characters interacting. The thread shows a sample video with Runway as well.
AI-Generated Images: Missed Opportunities and Moral Shrugs
Tom McDowell reflects on how much of our current AI image generation is very derivative and replicates specific artists. I agree with a lot of this. We could use these tools to create brand new and useful images, but there's a lot of use right now that feels problematic.
How People Are Really Using Gen AI in 2025
How do people use generative AI? I was surprised at how much of the use is for interactions like therapy and personal interaction. "Personal and professional support" was the top broad category in this research, which also includes tasks like organizing your life, resolving disputes, building lists, etc. "Content Creation and Editing" was second, followed by "Learning and Education." The popularity of learning on the list was also an interesting finding.
“Personal and Professional Support” is now the largest theme by far, stealing most of its new ground from “Technical Assistance & Troubleshooting.”
The power of generative marketing: Can generative AI create superhuman visual marketing content?
This abstract of a marketing research paper explains how they compared AI-generated images and ads to human-created ads. The AI images were not just comparable, they were better than what people created. While this research is specific to marketing, I think it's relevant to images we use in training and elearning.
First, we prompt seven state-of-the-art generative text-to-image models (DALL-E 3, Midjourney v6, Firefly 2, Imagen 2, Imagine, Realistic Vision, and Stable Diffusion XL Turbo) to create 10,320 synthetic marketing images, using 2,400 real-world, human-made images as input. 254,400 human evaluations of these images show that AI-generated marketing imagery can surpass human-made images in quality, realism, and aesthetics. Second, we give identical creative briefings to commissioned human freelancers and the AI models, showing that the best synthetic images also excel in ad creativity, ad attitudes, and prompt following.
When AI and Disability Rights Collide: A New Frontier in Workplace Accommodation 🤖♿
This article is about accessibility and AI tools in the workplace, but it's good to keep in mind as we experiment with and evaluate AI tools for training purposes too.
"5 Ways AI Tools Create New Barriers
1️⃣ Inaccessible interfaces – Not compatible with screen readers or voice recognition software.
2️⃣ Rigid productivity metrics – Penalize neurodiverse work patterns.
3️⃣ Communication analysis – Misreads communication styles tied to certain disabilities.
4️⃣ Cognitive assumptions – Built around neurotypical thinking.
5️⃣ Sensory requirements – Require inputs (voice/vision) not accessible to all. "
OpenAI.fm
Test out OpenAI's voices, available through their API. The "vibe" option lets you describe the tone, voice, and delivery. These voices do better at emotions than many of the other AI voices.
The cost is lower than ElevenLabs, but you need to use the API for actual production. I assume they'll create a web interface for use at some point in the future. OpenAI is vague about the data used to train its voices, which I assume means they didn't pay to license voice actors like ElevenLabs does.
LanguaTalk: learn languages with 5-star tutors & AI (Langua) • LanguaTalk
AI-powered audio chats for practicing speaking as you learn a language. The system has built in corrections and tips to help you progress. This use of AI makes a ton of sense to me. A human tutor would still be better, but this would be great for practice on your own.
illustration.app - AI Vector Illustration Generator
I haven't tested it out yet, but this looks like another viable option for creating vector illustrations in consistent styles. One feature I haven't seen elsewhere is the color palette generator. Give it a word like desert, nature, or neon, and it will generate a color palette that you can use in your images. You can edit and recolor images or make the background transparent directly in this tool.
How to Master Composition in MidJourney: Camera Angles, Poses, and Optical Tricks
Expanding your language to describe images helps you prompt better in AI image generation tools like Midjourney.
Create Consistent Characters within ComfyUI
ComfyUI is an AI tool for creating images, videos, and animations. This page talks about the workflow for creating consistent characters for stories.