By reading online, I sign the resources that I find interesting and useful. I share these links about once a month here on my blog. This article includes links related to the generation of IA images, images of character, AI voice and automated podcast transcriptions.
I generated the images below in Midjourney by retalizing a flat icon (also generated in Midjourney). Midjourney can keep the structure of the form but provide many variations in textures, materials and styles depending on your prompts.
Generate AI images
Suggestions and examples for the use of images generated by AI significantly to support learning, without adding confusing or distracting images. Consider the 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 the essential cognitive treatment of basic concepts. For example, a diagram illustrating the five forces to wear can help students better understand this business strategy framework.
- Decorative images: These graphics improve aesthetics but do not influence learning. For example, the image of a corporate handshake can be visually attractive but will not support or do not obscure the understanding of student negotiation strategies.
- Distracting images: Sung and Mayer call this category of “attractive” images. Although these visuals can be linked to the subject, they hinder learning because they require foreign cognitive treatment. For example, consider a complex organizational table of a complete company in a lesson on the leadership of the team. The image is largely connected to the lesson, but also highlights many non -relevant details, distracting students from key concepts.
The recent history of AI in 32 otters Ethan Mollick shows the progress of the IA image and video generation with iterations of an prompt on otters using wifi on an airplane. He also explains the difference between broadcasting and multimodal image generation models (Midjourney vs chatgpt). These tools get so different results because the technology and the underlying approach are different.
While the LLMS generate a word from the text at a time, which was always forward, the diffusion models begin with a random static and transform the whole image simultaneously through dozens of stages. It is like the difference between writing a sentence of history by sentence compared to start with a block of marble and gradually sculpt it in a statue, each part of the image is refined at the same time, not sequentially constructed.
But what makes the diffusion models interesting is not their growing capacity to make photorealistic images, but rather the fact that they can create images in various styles.
Unlike the diffusion models that transform noise into images, the multimodal generation allows large language models to create images directly by adding tiny colored patches one after the other, just as they add words one after the other. This gives an in -depth control on the images he creates.
IA image generators tend to exaggerate stereotypes
The examples of this article all come from older images, but the problems of biases in the IA image generators remain. Unless you explicitly worry about avoiding stereotypes, AI image generators reflect the bias of the images on which they have formed. Even if you invite you to avoid stereotypes, it can always be a problem.
Demystify the loras: what are they and how are they used in stable dissemination?
Quick introduction to Loras and how they are used for the generation of fine adjustment images in different styles for stable diffusion (and other image generators)
Character image generation
Creation with Gen-4 image references
Runway Gen-4 offers a more controlled workflow to generate images with coherent characters and scenes. Add a unique reference image of a character or more reference images to combine several characters in a scene or specify the parameter or objects. It is a more complex work flow than simply chatting with Chatgpt, but this gives you more precision and more consistent results. This is Runway documentation on the use of image references. My preliminary tests with Runway were very promising.
How to achieve the consistency of the character
How video of Flora on how to create images with the consistency of the characters in different scenes. It is a longer and technical process involving the formation of a LORA (low -rank adaptation) on an initial set of images for a character. It probably works better with real people, but there can be ways to adapt this workflow for Elearning with generated characters. This is more effort than what I would do for most projects, but it can be worth explored if I need something superior to a specific project.
AI voice
The progress of the voices of AI continue to impress me. Rime is more intended for organizations using AI voices for customer service or live conversations, it could therefore be useful for vocal chat in training applications. There is a free plan available to test it.
Podcast Automated Transcription
Podcast Transcript AI – Transcribing any podcast for free!
Generate a transcription of any podcast on Spotify or Apple podcasts. Look for the name of the podcast, choose an episode and receive a transcription by e-mail. I wanted to get a transcription of one of my podcast interviews, and it looks like a quick way to generate one.
Additional organized resources
Consult my full links library Or my previous Post bookmarks.
Related
At Learnopoly, Finn has championed a mission to deliver unbiased, in-depth reviews of online courses that empower learners to make well-informed decisions. With over a decade of experience in financial services, he has honed his expertise in strategic partnerships and business development, cultivating both a sharp analytical perspective and a collaborative spirit. A lifelong learner, Finn’s commitment to creating a trusted guide for online education was ignited by a frustrating encounter with biased course reviews.