📰 TL;DR

The implications of AI for misinformation and fake news have never been greater. This week, an image of Donald Trump surfaced on Twitter purporting to show his arrest. So lifelike, many people took it as real. The tweet received 5.6m views and nearly 5000 retweets. But, it was a fake - generated by an AI program that you can learn to use in about 5 minutes.

Sam Altman (OpenAI’s CEO) has stated that AI misinformation is his chief concern. I can see why.

[ 5 min read ]

Deep Dive 🤿

• What is Midjourney and how do I generate images from it?

• How to prompt Midjourney

• Future Implications

POV: You just talked smack about Will Smith's wife at the Oscars

This was created in Midjourney V5. The Prompt was: gopro view of Will Smith wearing a tuxdeo slapping the camera forcefully, Oscars award ceremony in the background, fine art cinematic portrait photography, ultra hyper realism, dramatic lighting, action photograph --v 5 --s 750

So was this. Prompt: photo of a cuber goth girl with skin made of blue and white porcelain --v 5 --ar 9:16

And this. Interior of a rustic kitchen with exposed wooden beams, fine details, super Wide Angle, real estate photoshoting. --no people

The details are outstanding. Whilst you can’t use V5 unless you pay, you can use V4 for free which is still crazy good. Want to learn how?

What is Midjourney and how do I generate images from it?

Midjourney is an AI program developed by an independent research lab called Midjourney, Inc., in March 2022. This program generates images and art from text-based prompts. According to their website, Midjourney is an AI project with a self-funded team focussed on design, human infrastructure, and AI, exploring new mediums of thought and expanding the imaginative powers of humanity.

The team consists of 11 full-time staff and an impressive set of advisors.

While similar to ChatGPT's sister-product, Dall-E, Midjourney distinguishes itself by the quality of images it produces. Currently Midjourney generates more complex and detailed images than Dall-E or Stable Diffusion (it’s AI Image Generator competitors).

How do you create a MidJourney account?

Signing up for Midjourney is similar to most sign-ups, but the process does not take place directly on the Midjourney website. Instead, Midjourney uses Discord as its sign-up platform. Discord is a free community-based chat application, similar to WhatsApp. If you want help understanding how to use Discord, you can refer to this article.

To join Midjourney's beta program, first visit the Midjourney website and click on the 'Join Beta' button located at the bottom of the page. Alternatively, you can click on this link to access MidJourney directly and skip to accepting the invite.

Once you’ve clicked through. Accept the invite to Discord (sign up to discord if you need to) and login.

Congratulations, you have successfully joined the MidJourney server! I understand that finding what to do can be a little confusing at first, but with a few clicks, it will all start to make sense.

When logged in, your screen should look like the image below. You can scroll through other peoples’ images in the chat.

How to prompt Midjourney

To get started, you need to navigate to a 'room' on the left-hand side of the MidJourney server called 'Newbies'. The 'Newbies' room will be located under the 'Newcomer Rooms' title. You can choose whichever 'Newbies' server you want to join; it doesn't matter which one.

In this example, let's select 'Newbies-111' room. Once you have entered the correct room, you will see a chat-like function with a message box. You will also see images that have been generated in front of you within the chat. You can read what people have prompted above their images.

To create your first MidJourney image, you will need to use a Discord command. Don't worry; this isn’t complicated. All you need to do is type '/imagine' in the text box.

Click on the /imagine command provided by MidJourney bot and type in your prompt.

So, type in the prompt (or text description) you want Midjourney to create. For example, in this instance I will type in “A cat flying through the countryside, pixel art”. Don’t ask me why.

Midjourney will now take some time to render your image. Usually around 20-60 seconds total, depending on how many tasks are running within Midjourney at that particular time.

It's important to note that when you start using Midjourney, you will only have access to the free trial version, which allows for approximately 25-30 image generations. After that, the system will prompt you to sign up for the paid version. Don't worry, this is more than enough to have some fun and create what you want. If you want to check how far you are into your trial, type '/info' into the chat/text box.

How to find your art and images within MidJourney

Once you've pressed enter and are waiting for your image to generate, don't panic if you can't find it. Just click on the 'inbox' section of Discord, which will show a little red notification once the image has been generated.

You’ll then see that the related MidJourney image that has been generated for you in the mentions folder.

Hover over the right-hand-side of the image and click ‘jump’ and it will take you to the message within the chat. As there are so many people using Midjourney, your image can usually get lost in the chat with other peoples images, that’s why it’s usually easier to do it this way.

Upscaling your MidJourney Images

MidJourney creates four images for you based on your initial prompt. These are essentially four different randomisations of the rendered image. You can simply click on the image and save it. However, if you want to explore get more out if it I’d continue.

In the image above, the U boxes, refer to ‘Upscale’ and the V boxes refer to ‘Variation’.

When you click on the U boxes, you’re choosing to upscale the image. This will enlarge it and improve the quality so that you can download it there and then. It will only provide you with the download for one single image that you selected.

On the other hand, when you click on the V boxes, you’re choosing to create different variations of the corresponding image you selected. MidJourney will generate 4 different versions of the image for you to choose from.

The circular arrow button is used to run the command again and generate a new set of images based on the same prompt.

To upscale image one, select U1. To generate different variations of image three, select V3. You got it?

Go ahead and try it out by typing in a prompt in the text box. If you're having trouble coming up with a suitable prompt, you can visit a website like PromptHero for inspiration.

In PromptHero you can peruse other peoples ‘artwork’ and copy their prompts. This is a great way to search for the kind of image you’re looking for without wasting your time going back and forth.

As with most AI programs and models, it’s important to spend time crafting your prompt. Minor changes in your prompt such as “A cat and a hat” and “A cat with a hat” can have significant differences in the resulting artwork. If you want to improve your prompt engineering skills, you can check out our community learning at Promptstacks. After logging in, navigate to the Text-To-Image School section.

Adding modifiers and differentiators to your text-to-image prompts

In the meantime, you can add different modifiers and styles to the end of your prompt to get different results.

“hyper realistic, 4K, 8k, HD, photorealistic, epic composition, Unreal Engine, Cinematic, Color Grading, Ultra-Wide Angle, hyper-detailed, insane details, intricate details, beautifully color graded, Unreal Engine, Cinematic, Color Grading, Editorial Photography, Photography, Depth of Field, Tilt Blur, White Balance, 32k, Super-Resolution, Megapixel, ProPhoto RGB, VR, Halfrear Lighting, Backlight, Natural Lighting, Incandescent, Optical Fiber, Cinematic Lighting, Studio Lighting, Soft Lighting, Volumetric, Contre-Jour, Beautiful Lighting, Accent Lighting, Global Illumination, Screen Space Global Illumination, Ray Tracing Global Illumination, Optics, Scattering, Glowing, Shadows, Rough, Shimmering, Ray Tracing Reflections, Lumen Reflections, Screen Space Reflections, Diffraction Grading, Chromatic Aberration, GB Displacement, Scan Lines, Ray Traced, Ray Tracing Ambient Occlusion, Anti-Aliasing, FKAA, TXAA, RTX, SSAO, Shaders, OpenGL-Shaders, GLSL-Shaders, Post Processing, Post-Production, Cel Shading, Tone Mapping, CGI, VFX, SFX, insanely detailed and intricate, hypermaximalist, hyper realistic, super detailed, photography, cinematic, intense, cinematic composition + intricate detailed, cinematic lighting + rim lighting + color grading, —v 4”

By adding a few of these keywords to your prompt, you’ll render really interesting images. Make sure to write out the prompt such as “A cat with a hat” and then the keywords like so.

“A cat with a hat, photorealistic, 4k, insane details, unreal engine”

Future Implications

The quality increase between Midjourney V4 and Midjourney V5 is outstanding. Considering V5 was only released 5 months after V4 and looks almost life like, you can imagine the implications this has for a range of industries. Take the film industry. Generating artwork through CGI has always been an intensive and expensive process. Now, or at least within the next year - anyone can create a extremely high quality image of choice in about 3 minutes. As we’ve said before, this could have major implications for fake news, deep fakes and misinformation. Only time will tell if this is an industry needing intense regulation. But what we can be certain of, is a rapidly changing landscape.

Editors Product Picks 📺

Rask AI

Rask is an AI-powered transation tool for videos. It provides automated voiceover, captions, and translation services. Rask can quickly localise video content allowing businesses of all sizes to create multi-lingual, dubbed videos. Oh and it’s currently got a 14 day free trial running.

LeiaPix

LeiaPix is really cool. Essentially, it allows you to upload a still image and make it 3D. It does this by adding depth through AI generated background and space.

Here’s a very low quality example below. Generated from our image at the beginning.

HeyDay

Basically, it’s an AI powered memory assistant. Super helpful when you’re researching.

Reddit Spotlight 🔦

ChatGPT now supports plugins. So, you can add processes and data to it even easier!