Midjourney, What is it, how to use it, and its limitations.

Midjourney, What is it, how to use it, and its limitations.

Image Generated using Midjourney

I recently published this article, where I am talking about some MidJourney tricks and tips after which I got some messages saying what I Midjourney. I know my bad, I did not talk about Midjourney a lot in that article and just assumed that you might be knowing about it and have an account set up already, but no worries, I will try to cover and address your queries in this Article, so let’s get started.

What is Midjourney?

In simple terms, Midjourney is a Text-to-Image generation AI tool.

What does it mean?

It simply means you can give a text prompt to Midjourney and it will respond with an image based on the given prompt.
The prompt can be anything*, existing, non-existing, living or non-living things, planets, galaxies, micro-organisms anything*.
Even though I said, anything, you can not generate anything, Midjourney is a censored platform and it returns with error messages for some prompts, and repeating such prompts may lead to the dismissal of your account.
The other day, I asked Midjourney to generate an image of a Car Crash and it responded with an error message saying the “car and crash” are illegal prompts for Midjourney.


The Prompt: Alien invasion realistic Result:

Image generated by Midjourney, Prompt: Alien invasion realistic

The Prompt: A retro diner, isometric, 3d, blender, night, raining, detailed Result:

Image generated by Midjourney, Prompt: A retro diner, isometric, 3d, blender, night, raining, detailed

How to use it?

Setting up Midjourney:

Simply register here: https://www.midjourney.com/home/, it will redirect you to Discord where you can sign-in using your Discord account and Midjourney will be added as a Discord Bot where you can experiment with different prompts.

After setting up, you can simply provide the prompt/instructions based on which it will give you 4 different options like this:

Midjourney options for Prompt: A retro diner, isometric, 3d, blender, night, raining, detailed

Where you have options to
U: Stands for Upscale (U1: Upscale the first image, U2: Upscale the second image and so on)
V: Stands for generating variations (V1: Generate variations using the reference of the first image and so on)
Refresh: Regenerate/Refresh the results (in case you are not happy with the current results).

Apart from simple text-to-image, Midjourney also allows us to describe or blend images, let’s understand these options.

The describe command:
Using the describe command, you can ask Midjourney to describe what it sees in the image. It simply means you upload an image and Midjourney will tell you what it sees in the image.
I asked Midjourney to describe the following image:

Image generated using Bluewillow

and it responded with the following responses:

Midjourney /describe command responses.

The blend command
This command allows us to blend multiple (up to 5) images using Midjourney like this:

I asked Midjourney to blend the Alien Invasion and Retro Diner images from above and here is the result:

Midjourney Blend Result

I know, it might not be the expected result but we can try refreshing it a couple of times to get the desired result.

Possibilities of Midjourney:

Midjourney is extremely powerful, people are generating paintings, artwork, app icons, illustrations, social media posts, images/banners for Websites, landing pages, product images, and even app designs using Midjourney.

Here are some examples:

Prompt: Flat illustration, Delivery boy in yellow, riding a yellow motor, white background, medium shot, 8K

Prompt: magical magic shroom encyclopaedia stefanie dahle style watercolour painting fantasy illustration style style

Prompt: Camping boy, camping girl, wavy curly hair, cartoon dessign, carrying camping equipment, wearing cute hats, spring green grass, tent undder trees, gradient style, sunny day, full body, trendy blind box clean background, natural lighting, 2k, super details, 3D, depth of field, Pixar trend, super realiism, ray tracing, complex details, fine textures, blender.

Prompt: digital oil, 1937 Bugatti, whimsical and imaginative, soft snowfall, pastel pinks, blues, and whites, icy lake with snow-covered palm trees, merging tropical and winter themes, JeeYoung Lee

You can check the Midjourney community showcase for more images and prompts.

Apart from these, Midjourney understands the terms of Photography, where we can mention the kind of picture we want to generate, Animated, Realistic, Paper layered, 3D, and 4K.

Limitations of Midjourney:

  • Censored: As I mentioned earlier, Midjourney is a censored platform, it can refuse to generate images for certain prompts and can block your account if you are continuously providing such prompts.
  • Context understanding and Bizarre results: Midjourney is an AI platform and sometimes it does not understand the context provided by the user, for example, I gave Mijourney a simple prompt: “hot dog” and here is the result generated by Midjourney:

Prompt: Hot dog

Images 1, 2, and 4 are fine but what’s happening with image 3, it’s a hybrid of Human, Hot Dog, and Dog and I am unsure as to why it created an image like this.

  • Text Limitations: Midjourney is an image generation platform (and not a language model) as a result, it does not understand text, hence if you ask it to generate a logo with some text, it will generate the logo properly but it will mess up the text.
    Prompt: generate logo of “Future of AI”

Prompt: Future of AI

The generated images are wonderful but the text next to random since it does not understand the text context.

I hope this article was helpful for everyone ad since now you know Midjourney, you can check out this article where I have shared some Midjourney tricks which might be helpful to you.

I hope you enjoyed reading this article, You may follow me on Twitter, LinkedIn, or right here on Medium where I am active and share similar content.