2024 was a relatively quiet year for AI image generation, at least compared to the high-profile years of 2022 and 2023. Video generation has become a hot topic, boosted by high-profile releases like OpenAI’s Sora.
“A lot of tools are trying to figure out where AI content production is going, trying to stay current, and diversifying into,” said Ross Simmons, co-founder of creative AI company ZenRobot. I think so,” he says. “But we’ve found that tools like Midjourney that stick to what they do best prevail.”
Most modern AI image generators continue to use diffusion models. This type of AI model is trained by training a neural network to predict random noise that is gradually added to an image. Once trained, you can reverse the process and create your own images from random noise. The diffusion model is then paired with a text model, converting the user’s text prompts into tokens that drive the diffusion model’s results.
While the adoption model has improved since 2022, it remains fundamentally the same, and over the past year the spotlight has shifted from the model itself to the software tools that underpin it. Today, the best AI image generators are fast, easy to use, and offer more control over image generation.
Midjourney is clearly popular for professional image generation. Midjourney
There are literally dozens of image generators available, but Midjourney is clearly the top choice.
“Midjourney is a very important tool for us,” says Chrissy Cremers, co-founder of Igency Amsterdam. “We’re still very good at art direction,” Simmons agrees, noting that he uses Midjourney “primarily.”
Released in 2022, Midjourney quickly gained a reputation for its detailed and captivating images with a strong artist style. They tend to produce images with bright, bold colors and a dramatic look that grabs the viewer’s attention. Midjourney’s dissemination model is unique and only available if you subscribe to its titular online service.
Creative professionals love Midjourney’s user interface, which is a bit ironic considering the tool’s origins. This model was initially only available through a Discord chatbot connected to an integrated user authentication system. However, things quickly changed, and Midjourney now offers a beautiful web interface that sets it apart from other products.
The most important new model released in 2024, Flux, can support image editing tools. Flux
While Midjourney remains a favorite among AI-powered creatives, 2024 saw the release of one major pervasive model that takes on a different role: Flux. The model was created by Black Forest Labs is a company founded by three former employees of Stability AI (the company behind Stable Diffusion).
“In terms of realism, this is amazing,” Simmons said, citing Flux as the most notable new image generation model of 2024. Kremers agreed, saying it was “much better suited for photorealism than Midjourney.” Flux also provides powerful tools for “in-painting and out-painting.” This allows the model to edit the image by removing unnecessary objects or expanding the image beyond its original extent.
In contrast to Midjourney, Flux is an open model. That means it’s free for anyone to get, use, and tweak. You can install it locally and generate images on your powerful home or office computer.
Leonardo.ai has carved out a niche by targeting creative professionals with specific needs. Leonardo.ai / Fyfe Boyce Design
Founded in 2022, Leonardo.ai has quickly established itself as a leading choice for individuals and businesses seeking professional image generation.
The company primarily focuses on creating tools that use existing generative models in new ways, but in 2024 it released its first in-house foundational model, Phoenix. Leonardo.ai also offers Flux, which ranks as one of the better options for these users. You cannot run Flux on your own hardware.
“I think it’s good if you’re generating illustrations or fantasy space or 3D characters to keep the characters consistent,” Simmons says. This is achieved through support for Low Rank Adaptation (LoRA), which allows users to tune the diffusion model to suit their needs.
Dall-E 3, available through OpenAI’s ChatGPT, features image editing tools to add or remove objects from generated images. OpenAI
OpenAI’s Dall-E was one of the first image generators to make headlines. But in 2024, it was overshadowed by the hype surrounding OpenAI’s video generation tool Sora, which was finally released in December.
Despite this, Dall-E remains one of the more prominent tools, especially for amateur use. Now part of ChatGPT, Dall-E quickly generates passable images. The April 2024 update also added image editing to allow users to remove or change specific areas of images generated by Dall-E.
Although Google’s ImageFX is just an “experiment,” it produces realistic and detailed images. Google
Google Labs, the home of the company’s “AI experiments,” includes an image generator called ImageFX. Although ImageFX is not widely promoted and has minimal features compared to other tools on this list, it has gained a reputation among savvy AI creatives.
“It’s a bit of a dark horse. It’s not really talked about, but it’s a good tool. It’s free, it can generate four images at a time, and the quality of the images is very good,” Simmons said. Ta.
ImageFX has a simple interface that prompts the user to enter a short prompt. It also provides a list of keywords to direct the user’s prompt to text that the model can easily understand. This approach allows you to easily generate detailed images with a specific style. ImageFX is also completely free, which is a plus for more casual users.
best of the rest
While the image generators mentioned above stole the show in 2024, they’re not the only tools worth paying attention to. Magnific has a reputation for excellent image upscaling. Krea.ai provides an attractive interface to third-party models, including Flux.
Creative professionals don’t seem to be too keen on image generators from established companies. Getty Images, Shutterstock, and Adobe offer generative AI services, but there is no buzz surrounding leaders like Midjourney and Flux. Kramer criticized the images used in the training. “They tend to be trained with a lot of stock images, and that’s reflected in the results.”
2025 will likely be a repeat of 2024. Video generation will continue to gain traction, as some companies (notably Google) have yet to release their long-awaited models. Still, Simmons believes creative professionals will continue to explore new ways to use image generation tools as they become more familiar with them.
“I wish people would understand more of the capabilities of AI. There are things it can do and things it can’t do,” Simmons said.
From an article on your site
Related articles on the web
