Head over to our on-demand library to view sessions from VB Transform 2023. Register Here
Midjourney, one of the startups at the bleeding edge of generative AI imagery, debuted a generative infill feature that users have already dubbed a potential game-changer, putting it in more direct competition with Adobe Photoshop’s Generative Fill.
An increasingly popular use of AI-powered image generation tools is to add entirely new elements within preexisting images, elements that fit the original image’s style. This is known variously as “inpainting,” “infilling” or “filling,” though the process behind each is largely similar.
More than the simple cut-and-paste that’s been around for decades, these features allow users to select a portion of their image and simply type a description of what they wish to add to their image in a text box. The AI automatically generates new objects or subjects, modifying the image by including them as though they had been there to begin with. It can also remove, replace or adjust what’s already in the image automatically, based on what a user types.
It’s an AI process that’s incredibly intuitive and efficient — and instantly speeds up and simplifies the image editing game.
VB Transform 2023 On-Demand
Did you miss a session from VB Transform 2023? Register to access the on-demand library for all of our featured sessions.
The infill feature that Midjourney introduced this week is called “vary (region).” It is a selector tool that lets you focus on specific areas of a Midjourney-generated image and “reroll” them for different results. Click on the “vary (region)” button on the Midjourney interface in Discord, and an editor window pops up with two tool buttons: a rectangular selection tool and a lasso selector. Use one to select the area you wish to modify and click “Submit” and the AI will attempt to automatically re-generate that specific part of the image.
You can also use text to generate a new region. Just type “/settings,” turn on “remix mode,” and voilà, you get a text box in the editor described above. You can now change the prompt for the selected region, giving you even more control and customization.
AI infilling: Not just for Adobe users anymore
Adobe’s Photoshop has been the go-to software for image editing for decades, and its Generative Fill feature, which debuted in May 2023, has been hailed by designers, illustrators and other AI users as one of the best around.
But with Midjourney’s new “vary (region)” feature, the landscape is changing. Users on social media are making direct comparisons and are already coming away impressed.
One drawback is that you’re not able to upload your own content yet, so you’re only able to use the infill tool on Midjourney-generated images.
In light of Photoshop’s yearly subscription fee of around $240, Midjourney is offering its users a more affordable way to leverage these new kinds of tools. Midjourney’s basic plan is just $10 a month, or $96 when you buy a whole year up front — though Midjourney offers a much more limited overall toolset than Photoshop, with no brushes, palette, layers and layer masking, etc. Meanwhile, Stable Diffusion inpainting is only for research purposes.
Now, there’s a bit of a learning curve here, but don’t worry, we’ve got some tips from Midjourney on how to best use the new features. Vary (region) works best on larger areas, about 20% to 50% of the image. For smaller tweaks, the “vary (subtle)” tool might do the trick instead. When changing prompts, keep it relevant to the context of the overall image.
“Changing the prompt will work best if it’s a change that’s more matched to that image (adding a hat on top of a character) versus something that’s extremely out of place (a dolphin in a forest),” read the Midjourney announcement.
AI design features seeing rapid evolution
It’s still early days for the prompt remixing tools. Midjourney says the tool can sometimes be unpredictable, generating outputs that are the opposite of what you’ve asked for, so take it slow and adapt as you go.
Vary (region) is just one of the new features Midjourney is adding to its portfolio. In June, Pan was introduced, allowing image creators to add context–aware fills to the border edges of the generated content.
More generally, generative AI in the design space is evolving on a regular basis. Last week, OpenAI announced its acquisition of Global Illumination Inc, a studio that builds “creative tools, infrastructure, and digital experiences,” according to an OpenAI blog post. As well, Modyfi, a browser-based image editor with powerful AI-enabled tools, entered public beta while announcing a $7 million round of funding.
In a nutshell, Midjourney has added an affordable option to the creative software landscape with its AI infill feature, vary (region). Giving Adobe Photoshop’s Generative Fill some good competition, it’s offering users a fresh, fun way to remix their generative imagery.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.