Technology Advisor Update – Spring 2024
The Artistry of AI Image Generators
The world has been all abuzz with talk of the increasing use of artificial intelligence, AI. While the field of AI has been around for many years, with the famous computer scientist and mathematician, Alan Turing writing about the “imitation game” in 1950. This later became known as the Turing test – a test to establish if a machine could be so good at mimicking human responses that, you can’t tell if you are interacting with a human or a computer.
AI has been quietly making its way into our lives. Apple’s Neural Engine, a form of AI was first introduced in the A1 Bionic chip found in the iPhone 8 in 2017. Today AI is found in mobile and other devices we use without even thinking about it.
However, it has only been since the wider release of OpenAI’s ChatGPT in 2022 that AI has come to the foreground of our awareness. We have since seen a plethora of AIs emerge, some using OpenAI, some using their own.
Despite some fears and concerns AIs can being very useful and fun to interact with. In this article I will focus on some generative AIs used to create images.
A generative AI, as the name suggests, using its trained model to create something, based on the data it’s given, in the case of the AIs listed in this article, some descriptive text data.
Midjourney
Midjourney quickly earned a reputation for producing rich coherent interesting and visually appealing images. Initially you were able to use Midjourney for free, however the free trail option is currently suspended. You now require a subscription to use the service. Subscriptions start from $10 a month or $8 a month for an annual subscription. This equates to being able to generate approximately 200 images a month.
Currently, you can only interact with Midjourney through Discord, making the interface a little tricky to use. You generate an image by typing the prompt /imagine followed by a description of whatever you would like to create. The AI will then generate four images you can then choose to either download, upscale, or re-edit, the image.
Your generated images are public, so can be viewed by anyone who is connected to Midjourney’s Discord server. People can also view which images you have created by looking at your profile. This, and the fact that you can access other Discord servers is something to be mindful of there are eSafety concerns.
Midjourney are currently testing a web app which means the AI image generator will soon be easier to access and use.
Midjourney has a really good guide explaining how to get started as well as a guide on the more advanced features, such as different model versions, using different parameters, upscaling blending multiple images. Once you are familiar with using Midjourney, you can truly produce some amazing images. See the Midjourney website for more information.
DALL·E 3
OpenAI’s DALL·E 3 is perhaps the biggest and most popular AI image generator. DALL·E 3, is a significant improvement on the popular previous version, DALL·E 2. DALL·E 3 uses ChatGPT-4’s understanding of language to expand your prompts and as a result it produces more interesting, realistic, and consistent results.
The biggest advantage to using DALL·E 3 is it’s easy to use, particularly if you are familiar with ChatGPT. Currently DALL·E 3 is only available to ChatGPT Plus subscribers. A subscription starts from $20 a month. However, you can access DALL·E 3 for free if you have a Microsoft account and use Microsoft’s Image Creator. See the DALL·E 3 web page for more information.
Microsoft’s Image Creator
Microsoft Designer is a feature rich AI-powered graphic design tool. Currently, Microsoft Designer is in preview (i.e. testing phase) and is free to use. Microsoft states that once Microsoft Designer is officially released it will remain free, but more advanced features would then require a Microsoft 365 subscription.
Within Microsoft Designer is Image Creator, this uses OpenAI’s DALL-E 3 to generate images. The Image Creator is incredibly simple to use and produces images for each prompt you enter. Depending on the current server load, it can take some time for the images to be generated. These images can then be either downloaded, edited or opened in Microsoft Designer to create things using the image. See the Microsoft’s Image Creator for more information.
DALL-E 3 is also available through Microsoft’s AI chatbot, Copilot. Simply ask Copilot to create an image for you. You will, however, need to be signed into a Microsoft account to do this.
DreamStudio
DreamStudio is a powerful image generating AI. It is the official Stable Diffusion web app. DreamStudio allows you to enter various parameters for the images you would like to create. While this gives you greater control, some people may find the additional options confusing. DreamStudio do however provide a good user guide. Currently, DreamStudio requires you to purchase credits in order to generate images. However, you are given 25 credits for free when you create an account which is enough to get a good sense of what DreamStudio is like. See the DreamStudio website for more information.
Adobe Firefly
Adobe is a company that has quietly been working on AI for more than a decade. This is evident with their image generator AI, Firefly. Firefly can be accessed through a web browser. However, it is also making its way into Adobe’s products like Photoshop.
A lot of the current generative image AIs struggle with text generation (i.e. text within an image). Firefly however seems to cope really well with this making it a useful tool for creating images that need to include text, e.g. a logo.
Similar to DreamStudio, the Firefly interface is packed with options that stem from Adobe’s image creation and editing heritage, making it a truly powerful tool.
Adobe Firefly can be accessed through a free individual account that includes 25 credits a month. If you require more then there are various payment plans available, including discounted rates for Students and teachers. See the Adobe Firefly website for more information.
Google’s ImageFX
Google, a bit late to the AI image generator space have produced an AI that is capable of generating high-quality, realistic images, including objects that are difficult to render, such as hands.
Google’s ImageFX interface is filled with features, that make it easier to refine your prompts or generate new ones via dropdowns. ImageFX also provides style suggestions for example photorealistic, 35mm film, minimal, sketch, handmade, and more. This combination of features makes ImageFX perfect for beginners who want to experiment.
Google’s ImageFX is free to use, but does require a Google account. It can be accessed through a web browser at Google AI Kitchen. While you are there, be sure to check out MusicFX.
This article has only scratched the surface of image generating AIs. There are many more, such as WOMBO Dream; Pixray; StarryAI; Deep Dream Generator; Shutterstock AI Image Generator; Craiyon; NightCafe; and more.
As always, I would like to hear about how you are using mobile, and other technology, and AI too. If you would like to have a particular topic covered in the next newsletter, please let me know. Finally, please feel free to contact me if you have a question or need tech help and support.
Martin Pistorius
Karten Network Technology Advisor