March 1, 2024
Midjourney Alpha is here with AI image generations on the web


Are you ready to bring more awareness to your brand? Consider becoming a sponsor for The AI Impact Tour. Learn more about the opportunities here.


Midjourney, among the most popular image-generating AI services in the world today with more than 17.5 million users in its server on the separate messaging app Discord, has begun testing an “alpha” version of its website, midjourney.com, that will allow users to generate imagery directly there instead of using Discord.

The Midjourney Alpha website with an AI image generation field was first spotted yesterday by avid users such as Nick St. Pierre (@NickFloats), who shared screenshots and videos of it on X.

Midjourney today confirmed today in its Discord announcements channel that it had “begun testing an alpha version of the website which includes image creation,” however, the company noted the alpha release was limited to “people who have made >10,000 images in Midjourney.”

“Don’t worry, we will be making it available to more people over the coming month!” the announcement added.

VB Event

The AI Impact Tour

Connect with the enterprise AI community at VentureBeat’s AI Impact Tour coming to a city near you!

 


Learn More

For users to see how many generations they have and if they have passed the 10,000 image threshold, they can first log into Midjourney on Discord and type the slash command, “/info” into any of the channels or directly to the Midjourney Discord bot to receive an updated count.

Even though I only had 8383 images generated for my lifetime usage, and 5041 images generated in Midjourney’s slower, less GPU-intensive “relaxed” mode, I was still able to access the Midjourney Alpha website myself and test it.

What it looks like and how it works

The website’s interface is minimalistic and clean, with lots of “white space,” and launches with an “Explore” tab showing public generations from other users. There’s a blank rectangular text bar marked “Imagine” at the top with a plus, where eligible users can enter a prompt directly.

On the very right side of this bar is an icon of horizontal lines and circles representing dials, which the user can click on to launch a visual interface for adjusting their prompt settings.

This is a major departure from the current Discord version of Midjourney, which forces users to type various jargon text commands into their actual prompts.

So now, using the web interface, instead of typing “–ar 4:3” for a landscape image, the user can simply click on the “Landscape” option in this prompt settings menu and it automatically applies this setting, without adding any visible text to the user’s prompt.

Furthermore, using the buttons on the left side rail, the user can tap the “Photo” icon and see a reverse-chronological view of all their generations made on the service since they started, arranged in a grid and grouped by date. This view also shows the user’s current image generation jobs in progress.

Kneecapping the competition

In all, it bears a striking resemblance to Visual Electric, the rival image generation web service based on the open-source Stable Diffusion model and geared toward professional designers, which launched last week.

However, Midjourney has the advantage of having what many users consider to be a superior quality image generation model, and the Alpha website contains a tab marked “Likes” showing all the images the user has already previously “hearted” on Discord.

It further includes a prompt search bar below the image generation bar and to the right, where the user can type in a prompt keyword — say, “steampunk” — and see all the public imagery generated by other users with that prompt word.

There’s a “sun/moon” icon button in the lower left corner that the user can use to toggle between light and dark mode, and a “thumbs up” button that allows users to participate in a crowdsourced image ranking project to choose their preference between two images that the Midjourney team serves up — designed to help further improve the quality of their AI model.

However, one limitation I noticed is that the images generated using the web interface do not appear to repopulate in my Midjourney Discord Bot chat, meaning for users toggling back and forth, the web interface is the only one that retains all image generations from both sources.

Still, overall, it is an impressive alpha release and will no doubt make Midjourney more accessible to a wider potential audience. For now, though, you still need to sign in with your Discord username. Midjourney is also one of several AI image generation services and AI companies more generally facing lawsuits from artists and creators for alleged copyright infringement over its use of public-facing imagery to train its models.

More is on the way

An iOS app and folders on the web, according to Midjourney power user St. Pierre:

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.





Source link