Not that long ago we explained how you can make AI selfies with the Lensa app on your phone. The results can be stunning, but AI can go beyond using an existing photo as a reference point and create artwork from scratch.
There are now a lot of AI image generators which will produce images based on text descriptions. This means you can dream up anything at all, type it in and mere seconds later see a selection of images matching that description.
For example, the image at the top was generated using the description “steam train going across a curved viaduct on rolling green hills with a river running below, cloudy blue skies.”
But you can go into as much detail as you like. For example, the following produced these variations “A beach on a cloudy day. In the distance, people are walking towards cliffs in the background”.
Or even “1970s Porsche racing car on track leading a pack of period cars with crowds of spectators in stands in the background”.
AI is good at generating images of people. For example “a 35 year old IT officer working at his desk, wearing glasses and a blue blazer” led to this:
These generators vary in quality and usually allow you to generate only a few images before you have to pay to use their services.
But it’s fun to do, and well worth trying out as there’s every chance you’ll be blown away by the results.
Regardless of which generator you decide to use, the process is the same: type a detailed description of what you want. The longer that description, the better.
Shorter descriptions can still generate incredible images, as demonstrated by the birds above, but they may not be quite what you wanted. Plus, AI isn’t human – duh! – so although some of the images appear flawless, others have clear errors, and AI struggles with text in particular.
Really, you also need to specify the style of image you want. That’s because the AI needs to understand whether you want a photorealistic images, a cartoon, a sketch or something else. If you don’t specify, it may choose at random.
Since the images are created by powerful servers, you can use these generators on whatever device you have: a phone, laptop, tablet or PC.
Some generators are available only as mobile apps, but others work in a web browser. Here are some to try out now.
- Midjourney – Runs on Discord
- Fotor – works in a browser
- Photoleap – Android and iPhone
- DALL-E – works in a browser
DALL-E is one of the best known, and is from the same people – OpenAI – that made ChatGPT. Unfortunately, its popularity means it’s all but impossible to use: we could barely create an account without the website falling over, presumably due to being overloaded.
However, Midjourney stands out from the crowd, partly because of the sheer quality of the images it creates, but also because you’re given 25 minutes of GPU time for free: you’re not restricted to a specific number of images.
The only snag is that it uses Discord, which can be a little bewildering to use if you’re not familiar with it. You’ll also need to create a Discord account if you don’t have one.
How to use Midjourney
Create a Discord account
Go to Discord’s website and click the white Log In button at the top right. Click the Register link at the bottom, below the blue Log In button. Enter the information and click Continue. Verify your email address.
Join the server
To use Midjourney, go to the website and click the Join the Beta button. You’ll see a prompt to open Discord. This may open the app, or open in a new web browser tab depending on which device you’re using.
When asked, click Join Midjourney and prove you’re not AI by clicking on the images as requested.
Find a Newbie room
In Discord, click on the Midjourney server from the list of servers on the left-hand side. If you’re new to Discord, it’ll be the only one there. The icon is a pencil drawing of a sailing boat on water.
In the list of channels, you should see NEWCOMER ROOMS. Click on one of these.
Generate an image
In the message box at the bottom type /imagine and press Enter. A prompt box will appear. Type into this any description you like and press Enter. Here we’ve typed ‘photorealistic close-up of a stained glass bird’
Wait for your image
These ‘NEWBIE ROOMS’ are shared with other people, so you’ll see their creations appearing. Wait for yours to appear which can take a minute or so. You’ll get a selection of four. You can then click the buttons below to get a variation on one of those four by clicking on V1, V2, V3 or V4 (they go clockwise) or get a larger version of one you like by clicking the corresponding U button.
Specify aspect ratio
If you don’t want a square image, you need to tell Midjourney the aspect ratio you want. Do this by adding -—ar 16:9 if you want a TV-sized image. You can specify any ratio, including the common ones such as 3:2, 4:3 and 3:4.
Find your images
It can be difficult to find the AI images generated for you in the list of hundreds, especially if you don’t sit there waiting for them to pop up. But there’s an easy way. Click on the Inbox icon (circled) near the top right of the Discord interface.
Click on Mentions
Click on Mentions and you’ll see every message you’ve been tagged in, and you can click the Jump button (which appears in a web browser when you hover over a specific mention) to go directly to that post in the feed.