AI-generated images have been around for many years but AI-powered image generation tools have exploded in popularity over the past year (Knight, 2023). So if you are interested in creating images using AI, how do you get started? When I posed this question to ChatGPT, Google Bard, Claude, and Bing Chat, the answer was some variation of "It depends on what you are trying to do."
Here is a breakdown of four of the more popular AI image generation tools.
Midjourney (Subscription Plan)
Midjourney is arguably the best tool for 99% of users because of the low barrier to entry and high ceiling. Text prompts to generate images can be a few words or a few paragraphs. Things can get complicated when you start adding "negative prompts" and "weights" but overall the Midjourney proprietary training model generates very detailed and accurate images. Midjourney operates on Discord, which can take some getting use to, but it is still accessible for most users. Another advantage of Discord is that you can view a gallery of what people have generated and what is being generating in real-time. This social/community aspect allows you to message and learn from those who are more knowledgeable about AI-generated content.
Stable Diffusion (Free)
Stable Diffusion is the main competitor to Midjourney. It is 100% free and open source. As a result, Stable Diffusion has a strong, global user community and it's AI model is the foundation for many paid AI-image tools. It also has many advanced customizable features not found in other tools on the market, as well thousands of custom training models.
The downside is that Stable Diffusion is more complex to use. It requires users to have an understanding of AI, as well as a powerful computer with an external Graphics Processing Unit (GPU).
Dream Studio (Free + Paid Tiers)
For those that want to use Stable Diffusion but do not have the AI knowledge or equipment, you can use DreamStudio. DreamStudio was developed by Stable Diffusion and runs directly on your internet browser. Users just need to enter a text prompt. The accuracy and speed at which images are generated are high since most AI image generation tools use Stable Diffusion's open source code.
DALL-E 2 (Free in Beta)
DALL-E 2 is made by OpenAI, the same company behind ChatGPT. DALL-E 2 images are not as high fidelity as those generated by Stable Diffusion or Midjourney. However, its strength lies in its simplicity. Moreover, making edits to generated images is simpler than "repainting" images. Since you can customize an image frame by frame, you can achieve the desired image without having to type in 2-3 pages of text prompt.
One reason DALL-E 2 is not as popular is that it will not generate realistic faces due to the rise of deepfakes. But some believe the technology behind DALL-E 2 is better than competitor tools.
If I could only recommend one AI-image generation tool it would be Midjourney. Although Midjourney has a cost, I believe the image quality, accessibility, and convenience are worth it. The Midjourney user gallery and text prompt library are also a huge plus.
References
Knight, W. (2023, January 12). Where the AI art boom came from—and where it’s going.
WIRED. https://www.wired.com/gallery/where-the-ai-art-boom-came-from-and-where-its-going/