Midjourney Launches AI Image Editor: How to Operate It

Halfway through the journeythe successful AI image generation startup, founded and run by former Magic Leap engineer David Holz, is delighting users with a up-to-date feature unveiled last night: AI image editing.

As many of Midjourney’s 20 million-plus users (including some of us at VentureBeat) probably know, Midjourney previously allowed users to upload their own photos collected off-site to its alpha web interface and/or a Discord server to serve as a reference point for the AI ​​image generator’s diffusion models – the latest being Midjourney 6.1. After receiving the uploaded reference image, Midjourney’s AI model is able to generate up-to-date images based on the file provided by the user.

However, this reference function didn’t actually make any changes to the source image – it just used it as a sort of loose starting point.

Now, with Midjourney’s up-to-date “Edit” feature, users can upload any image of their choosing and actually edit sections of it using AI, or change its style and texture from the source to something completely different, such as transforming an elderly photo into an anime – while keeping most of the themes and objects of the image and spatial relationships.

It even works on user-submitted doodles and hand-drawn drawings, turning your doodles into full works of art in seconds.

Midjourney has released a video demonstration showing how to exploit the up-to-date features, which we’ve embedded below:

VentureBeat uses Midjourney and other AI tools to create content for our website, social channels and other formats.

Please note that despite its popularity, Midjourney is one of several artificial intelligence companies that have been sued in a class action lawsuit filed by human artists for alleged copyright infringement by scraping human-made works without express consent, authorization, consent or remuneration for training their models. The case remains in court for now.

The Midjourney image editor appears to be restricted to only the latest AI model, Midjourney 6.1, which makes sense.

In message to the Midjourney Discord communityHolz wrote that: “All of this is very new and we want to give the community and moderation staff time to gently get used to it…”

As a result, the up-to-date Midjourney Editor feature is currently only available to users who have generated more than 10,000 photos with the service, those with an annual paid membership, and those who have subscribed to the service for a year or more.

However, if you meet these criteria, you can exploit the up-to-date Midjourney image editor by following the instructions below.

How to find and start using the Midjourney image editor

The up-to-date Midjourney image editor is available exclusively on the alpha web interface, available at alfa.midjourney.com.

Once logged in, an eligible user should see a up-to-date button along the menu on the left, about halfway down, with an icon of a diminutive pencil on a pad. Hovering your cursor will show it reads “Edit” (or the text will automatically be permanently displayed if your browser window is wide enough).

Clicking on this should bring up a up-to-date Editor screen which should present the user with two main options: “Edit from URL” and “Edit Uploaded Image”.

The latter requires the user to have the file saved on their computer, while the former can accept a wide range of images hosted on various websites such as Wikimedia Commons if the user simply pastes the correct link to the image hosted on the internet. For the purposes of this article, I provided the URL of the file concept car image below from Wikimedia Commons.

After uploading a copy of the file to Midjourney via the URL or the user’s own file repository, the image should appear in the center of the up-to-date editor screen as follows:

You’ll notice that in the menu on the left side of the internal sidebar there is a wide range of options and different buttons that users can select to modify an image in Midjourney 6.1, including “1. Erase”, which allows the user to erase and paint over parts of the image using artificial intelligence using a brush and text prompt, “2. Move/Resize,” which allows the user to move an image around the virtual canvas and expand its edges with up-to-date, matching AI images, and “3. Restore”, which is the opposite of Erase and allows the user to keep any portions of the source image that they accidentally painted over with the Erase brush.

The user can control the size of the brush using the slider in the left sidebar, as well as the “scale” of the image, zoom in and out, and the aspect ratio itself, using more presets below.

There is also a “Suggest a Prompt” button, which in Midjourney explains itself when you hover over the text. It is intended to assist the user generate suggestions describing the image they just uploaded – in case they want to change this prompt or exploit it to generate a completely up-to-date, similar image. The suggested tooltip text should automatically appear in the tooltip box/bar at the top of the screen.

Looking at our concept car example, I went ahead and used the Erase brush tool on the driver and the text input bar at the top of the Midjourney web interface to replace the driver with a “flaming skeleton.” After entering the text prompt in the top input bar/box, I hit the button labeled “Submit Edit” or type on the keyboard to apply the changes.

Similar to Midjourney’s raw image generator, the Editor automatically creates four versions for each text prompt – evident in the right sidebar under the “Send” button.

Here is the best result of my experiment:

The user can then choose to continue making up-to-date changes to the resulting image, scale it using Midjourney’s built-in scaling module using the button below, or download it as is.

Retexturing turns images into up-to-date adaptations in different styles

Additionally, astute readers and Midjourney users will notice that when you click the “Retexture” tab in the left sidebar, another set of Editor options is also available.

As Midjourney itself explains in the left sidebar when you click this option: “The retexture will change the content of the input image, trying to maintain the original structure. To get good results, avoid hints that are inconsistent with the overall structure of the image.”

As you’ll see in the screenshot I posted above, there’s a lot less going on on the Rexture screen than on the regular edit screen. Basically your only option is to exploit the tooltip text input bar/box at the top of the screen to specify what kind of retexturing you want to do to the source image you/user provided.

Once this is entered, the user can click “Upload Rexture” and viola, Midjourney will exploit AI to apply the up-to-date texture and adjust the image according to the user’s instructions, again generating four versions for him to choose from.

In my case, I tried many different styles including but not restricted to anime, cave painting, colored sand, grotesque slime, and cyberpunk styles. See some of the retexturing examples I received below. One note on my restricted testing so far – the retexturing feature seems to warp and remove some details from the resulting source image, as well as gender-swap subjects and add up-to-date, extraneous details. But that’s part of the fun of using Midjourney or other generative AI imaginative tools – see what the model spits out based on your input!

Sultry reception among AI imagers on X

The AI ​​image and art community on social network X has applauded the up-to-date editor of Midjourney – something that has been rumored for several weeks. Some of the leading AI developers have already tried it and published their examples, many of which are impressive. Here is a sample:

If you are a Midjourney user and meet the criteria above, log in and try it out! Let me know what you think: carl.franzen@venturebeat.com. Midjourney has also been open about its plans to launch a 3D or video editor, which could arrive later this year.