Midjourney has been one of the best AI image generators for years now, and it keeps getting better. Not only does the quality of its generative AI improve with every version, but it now has a website that’s enormously easier to use than the Discord-based interface that Midjourney previously relied on.

In this Midjourney guide, we aim to answer the main questions people are asking about the service in one simple article. We’ll also show you how to use both the Midjourney website and the older Discord interface.

What do I need to know about Midjourney?

Here’s a breakdown of the questions we answer about Midjourney in this guide. If you feel we’re missing any, please email [email protected] and we’ll consider it for our next update of this article.

How to use the Midjourney website to create AI images

Once you’ve registered and signed in with Midjourney, you may be asked to complete a training exercise. Here, you’ll be shown a succession of images and simply asked which you prefer: A or B. This helps inform the AI about the style of images you prefer and allows it to generate images accordingly. We’ll cover this more deeply in the How do I personalise images? section later in this article.

Once any training is complete, you can start generating images by clicking on the Create link on the left-hand side. You should see a prompt window at the top, where you can describe the image you want the AI to create. Hit the submit arrow and the AI will create four images for you to choose from.

If all you want is a very straightforward image, that simple text prompt might well suffice. But if you want more granular control over your images, click the Settings icon to the right of the prompt bar.

Here you can determine what size the image should be (landscape, portrait or a range of different aspect ratios), as well as changing parameters such as stylisation (how much of your preferred style it should add to the image), weirdness (how bizarre you want the images to look) and variety (how different the four options you’ll be shown should be from another).

For example, he ask Midjourney to generate images of “a horse galloping across fields” with weirdness and variety both set to zero, and we get four reasonably similar, predictable images to choose from:

Midjourney-generated images of horses galloping across fields

Whereas if we turn the weirdness and variety all the way up to maximum with the same prompt, the images are much more striking and unusual:

Midjourney-generated images of horses galloping across fields

Bumping up the weirdness settings also raises the potential for AI oddities. In the bottom-left image above, for instance, there’s a horned devil seemingly growing out of the back of the horse!

The settings menu also lets choose which version of Midjourney to use, as well as giving you the option to use the new draft mode, in which you can experiment with different ideas using a faster-to-generate “rough” that won’t burn through as many of your credits as a standard image. If you find an image you like in draft mode, you can improve its quality later by clicking the Enhance button.

How to use Midjourney on Discord to create AI images

Even though Midjourney now has its own website, you can still generate images via the Midjourney Discord,

You can visit any of the 20 General Image Gen threads in the Midjourney Discord and start generating images there. However, if you use these threads, it’s easy to lose track of the images you’re creating amongst all of the other images being generated by other users.

Instead, it’s much easier to generate images using Direct Messages with the Midjourney Bot. To do this:

  1. Click on the Direct Messages icon in the top-left corner of the Discord app
  2. Select Midjourney Bot from your list of contacts
  3. Type your image prompts into the field marked ‘Message @Midjourney Bot’ at the foot of the screen

Click here for our detailed guide on how to write text prompts into Midjourney.

How to increase resolution in Midjourney

If you’ve entered an image prompt into Midjourney, you’ll be given four different options to choose from. However, these images are generated at a relatively low resolution that isn’t ideal for using at anything above thumbnail size.

So how do you improve that resolution? If you’re using the Midjourney website, click on the image you’d like to improve, and then on the right-hand side you will see two different upscale buttons: subtle and creative (as shown above).

If you want the AI to upscale the image as close as it possibly can to the image you see now, click subtle. Choose the Creative option if you see AI oddities such as obscured faces or too many fingers, and this may correct the errors. You will also see buttons that let you vary the image here, if it’s close to what you want, but not quite there.

If you’re using Discord, you can press the U1 (top left), U2 (top right), U3 (bottom left) or U4 (bottom right) button beneath the results to increase the resolution of that particular image.

Midjourney guide - increase resolution

How to use reference photos in Midjourney

You can give Midjourney reference images, hinting what you’d like included in the generated images. This unlocks some of Midjourney’s more fun features. For example, you can upload a portrait of someone and ask Midjourney to turn it into a Pixar character, such as with my portrait below.

Note that when you upload a reference image, Midjourney won’t create an exact replica of it, but use it for inspiration. The results can be mixed and some trial-and-error is required.

To use a reference image on the website, click the picture icon to the left of the prompt bar and a menu will appear on to which you can drag and drop relevant images. There are three options here:

  • Image prompts allow you to upload one or more images that you want to use a guide for the AI. You should still include a text prompt to describe what you want the new image to look like, but if you’ve uploaded an image of an oak tree and ask for an image of “a man sat under a tree” it should use the supplied image as inspiration for the type of tree you’re looking for.
  • Style reference lets you choose an image that you want to mimic the artistic style of.
  • Omni-reference is where you should upload the image of a person or the form of an object that you want to generate an image of.

To use a reference image on Discord, you’ll either need a link to an image hosted on a website or to upload an image to your Discord chat and generate a link from there. To upload an image into Discord, simply drag the image from your computer into a direct message with the Midjourney Bot, or click the + button in the message field and select the image, as shown in the screenshot below:

Reference photos 3

Once the image has been uploaded, right-click on it and select copy link.

Now create an image prompt in the following format:

/imagine http://sampleimage.jpg text-prompt parameters

So, for example, to create the Pixar character above, the prompt might be something like:

/imagine http://Barryportrait.jpg as a Pixar character --ar 3:4

The “–ar 3:4” parameter at the end sets the aspect ratio of the generated image to 3:4.

How to personalise images in Midjourney

Midjourney moodboard menu

As mentioned earlier, one of the newer features added to the Midjourney website is the option to personalise images.

By default, Midjourney creates a “global” profile for you, based on the image preferences you showed during the training period. However, you can also click on the Personalise menu and create what Midjourney calls Moodboards of images of a certain style. Here, for example, we’ve got a Moodboard of “moody noir” images, which are black & white images of high contrast.

We can tell the AI to take inspiration from our global profile/Moodboards when generating new images. In the Create menu, click the P button to the right of the prompt window and select which profile/Moodboard you want the AI to use. Then, before generating an image, move the stylize setting to the far right. Now enter your prompt, and your image should adopt the look you’re hoping for.

Here, we enter the following prompt:

“A sports car turning a sharp corner on a country road, photo-realistic, wheel spin, gravel surface.”

As you can see, we make no mention of the image’s visual style, but because we’ve selected the moody noir Moodboard and cranked the stylize setting up, the AI knows what we want…

A Midjourney-generated image of a sports car driving down a country road

How to edit images with Midjourney

Another great new feature on the Midjourney website is the editor. Now, if something’s not quite right in an image you’ve generated, or even in one of your own photos, you can edit the image to put it right.

Here we’ve generated a “stay fit” graphic, but as you can see it’s come with some extraneous text that we don’t really want:

"Stay Fit" graphic with extra text

With the editor, we can click on the generated image and click the Edit button in the bottom-right corner. With the brush tool, we can draw over the areas of the graphic we don’t want, then click the Erase button.

Midjourney's editor screen

Now we click the Submit Edit button at the top of the screen, and all of that extra text is removed.

Cleaned Stay Fit graphic

The editor can also add elements to an image. We’ve got a moody shot of a badger on the prowl, but we want to add some lightning to the sky. We open the image in the editor, brush over the section of image where we want the lightning to appear, and then type “lightning” in the text prompt window.

Midjourney editor adding lightning to a photo of a badger walking across a field

Once we submit the edit, we get four options to choose from, including this image, which subtly matches the lightning to the tone of the existing image to make it look like it was part of the original.

Midjourney-generated image of a badger walking across a field with lightning in background

How much does Midjourney cost?

Basic PlanStandard PlanPro PlanMega Plan
Monthly sub cost$10$30$60$120
Annual sub cost$96 ($8 per month)$288 ($24 per month)$576 ($48 per month)$1,152 ($96 per month)
Fast GPU time3.3 hours per month15 hours per month30 hours per month60 hours per month
Relax GPU timeNot availableUnlimitedUnlimitedUnlimited
All the prices are in US dollars and correct as of 27 May 2025

Most people should start with the Basic Plan, which costs $10 per month if you subscribe on a month-by-month basis. It works out as $8 per month if you buy an annual subscription.

Fast GPU time indicates that you get instant access (or as close as possible) to a GPU. With an “average” image taking less than a minute, that gives you a minimum 200 images per month.

The Standard Plan is a big jump in price but means you get 15 hours (900+ images), plus access to Relax Mode. Here, jobs are placed in a queue and done when that’s possible.

How to cancel Midjourney subscription

If you’ve decided you’re not getting value for money from your Midjourney subscription, you can cancel. The cancellation will take effect from your next monthly or annual billing cycle, not immediately. So if your payment went out last week, for example, you’ll likely get around three weeks’ more service before you’re cut off if you pay monthly.

To cancel your Midjourney subscription:

  1. Visit the Midjourney website and log in, if required
  2. Click the Manage Sub link on the left-hand side
  3. Here you should see your renewal date, which shows when your next payment is due
  4. Click the Billing & Invoice details button
  5. On the following page, press Cancel Plan

How to use Midjourney for free

Midjourney doesn’t offer a free tier any longer. When the service first arrived, you were able to generate a few dozen images for free, but that has since been withdrawn.

It’s worth keeping an eye on Midjourney’s Twitter account, as it does (very occasionally) turn the free trials back on for short periods. Otherwise, it will cost you $10 to try the service for a month, but remember to cancel your subscription using the instructions above before the next renewal date.

This article was updated on 27 May 2025 to reflect the changes to the user interface.

About The Author

Avatar photo
Barry Collins

Barry has 25 years of experience working on national newspapers, websites and magazines. He was editor of PC Pro and is co-editor and co-owner of BigTechQuestion.com. He has published a number of articles on TechFinitive covering data, innovation and cybersecurity.

Read more from this author.

We take journalism seriously. To learn more on why you should trust us, head to our editorial guidelines page or meet our team.