EdTech Lounge

View Original

21CL BKK: AI is Not Taking Our Jobs, Quite the Opposite: Midjourney Generative Art

This is a third part to creating generative resources in the classroom with the use of artificial intelligence. In this post, we are looking at what Midjourney via Discord can do for us.

Within this post, there will be an expectation to look into signing up to platforms, using apps, perhaps paying for apps or services such as Midjourney pro or, finding free/freemium/ open source alternatives to work alongside the process here.

See this content in the original post

Some prerequisites:

  1. Discord. Free to use and sign up via app or web.

  2. Midjourney Discord server. Free to view and use for 25 (x 4 - each image is a panel of 4 images, 100 inn total). Option to pay $10 per month.

  3. Photoshop or similar. Photopea is a good web alternative.

  4. Canva: an option to 3 for editing in a basic way - cutting, cropping etc. + below

  5. Your Operating system’s image editor.

  6. Optional: Comic book app or creator. Perhaps Canva has non-premium content. Adobe Express has many.

  7. Optional: Storyboard templates.

  8. Topic or theme you are working on with children for setting and character design.

  9. Optional movie editing software such as iMovie, CN Movie Editor, Capcut or your choice.

At my school I run several ECAs. Part of these are animation ECAs that allow the children to get away from the humdrum curriculum constraints and let the children go to far reaches of their mined for wild ideas, stories and characters.

One day we are in the ECA and there was a brouhaha from the other side of the room where some children we getting more and more frustrated with the results a search was throwing up for them. I mean, all they wanted was ‘Lionel Messi slide tackling Buddha in the penalty area’ ‘Simba as a cub falling to his doom from a cliff in the African savannahs’ and ‘Alex from Minecraft riding an actual roller coaster’.

I dawned on me that the way children want to search coupled with the way they think search should behave is exactly the way Midjourney behaves. If you don’t know what Midjourney is then it’s like this search above and it kind of generates a version of this mental imagery as a selection of four options for you to either remix and get variations or ‘upscale’ the one you like. Much like the one below. This is part of 1940s comic book theme in the style of Al Williamson. The prompt here is: Al Williamson, comic book character, 1940s, film noir, rookie cop alone, no background, black and white, --ar 1:1

You’ll note that this is different to the children’s ideas and for good reason. In our 21CL BKK workshop we will be aiming for a type of consistent theme, character, setting object or product that we can use in our classes.

You may also note that there are some numbers at the end, a ‘style’ and specific instruction for the character I want. There are panels I want in my comic, so I ask for this too. This is only so I can get a flavour of what I might want to produce later on. A bit of inspiration or a type of mood board from the prompts.

Al Williamson,, comic book set in the 1940s, black and white, detective, film noir, action shot, 9 panels, color only yellow

Discord

To do this, we need to sign up to Discord. Create an account.

Then check out some AI servers.

Midjourney

Then find Midjourney here. Click and join. You can use the App, on your device or on the web.

Once you’re in Midjourney you have 20 general chats on the side. Choose one and look around. Scroll up and down to see that people are making. Look at the prompts. Some are long, very long or filled with added extras such —seed 1233040302 or —AR 16:9 or the like.

Explore

Either way see if you can work out what they do.

Here we can see the prompt. Simple and strange at the same time.

Let’s see what you get when you try the same prompt. As a rule, when you sign up, you are only give 25 free images in batches of four. You may want try this, or look at the showcase here to get a style you like.

Then we will get some ideas for a writing project, a creative arts project, animation, computer game deign idea or story prompts for your classes.

Might be good to write some prompts of your own in your phone, text or pen and paper. Then, drop them into a prompt creator like AI PRM if you would like to get some wild ideas and use ChatGPT to help out.

See the video below and the link here for the Chrome extension.

AI PRM - Prompt Creator in Chrome for ChatGPT

See this content in the original post

Create a Character

One of the first ideas you may have is character choice. For this example I am going to set up a series of four characters and use one as a main thread here. The reason for this is that we can end up creating comics as a series of characters, poses and settings. You can skip down to see the characters in the story I’ve made up.

The hard part, as you will discover, is getting consistent characters. But we can practice on this. If you are already a user of Midjourney, then you may want to gather some characters and download the images you want. Then, collect the links to them. If you want a video - go to the bottom of this post.

In the meantime, if you are a new user then please create an idea for a character in a story.

Younger Children’s Use of AI

As you may be thinking, how can I use this with younger children, it’s hard to without an adult or older students beyond 13 years old. If you would like to use this process with younger children in class, then perhaps Craiyon.com might help you.

From this prompt in Craiyon.com: al williamson style, black and white, 1940s, film noir, comic book character, rookie cop alone, full figure, standing facing front, no background, white paper, we get this result. Not quite the same but usable all the same as a storyboard or a starter character at least!

Making a Character

Let’s move on…

Things you need now are these:

  1. character idea: man, woman, beast etc.

  2. style: painting, comic book,

  3. Artist style: Dali, Yayoi Kusama, Roy Lichtenstein, Bauhaus, 17th century Flemish, 21st century realism etc.

  4. Colour choices.

  5. Expressions.

  6. Clothing: hat, cloak, chaps…

  7. Character description: stopped, fat, thin, laughing, jolly, morose etc.

  8. Background style: (below I have made a white backdrop as it’s a character sheet). Plain, busy, patterned, Fleur De Lys, Jungles, Desert, Frosty Tundra. Experiment, but for this exercise I am showing character only until later.

For this character introduction, we will only use a character sheet. The reason why is that we can download the one we want trim it in an editor and re-use it. The other reason here, is that we are training the generative part of the AI to understand what we want. This will allow us to keep this as consistent as possible and this is key if we are to make a series of character poses.

Once you have your idea, go to Midjounrney and in the sidebar, where it has #general-20 (any one is fine), and type: /imagine + your choices.

I typed this: /imagine woman comic book dali style red dress bouncing curly hair happy and jolly walking white background character sheet

Of course, you can bow out now and play with these selections if you are new to this. For those who have used this, then please carry on reading as I am going to explain the options we have and why we need to keep track of this as it develops.

Going Further

Above, we can see the part finished image with about 20% finished. You can see though that the character is clearly coming through with the clothing and shape on the white background.

The video below, you can see that I am using the MidJounrney app on Mac, I’ve chosen the #general-3 channel. In here you can see my prompt: /imagine woman comic book dali style red dress bouncing curly hair happy and jolly walking white background character sheet

The image is fine for an introduction and will set you off if you want to follow this line of characterisation. As I've mentioned above, I am making a comic, and I want to make a 1940s style comic, so this means 7-9 panels. As you can see above too, I have used MJ for inspiration. I also researched some prominent artists of the time and found that Al Williamson’s style is what I like. I like the story lines he was involved with and I will stick with this.

The Interface

You will notice that there are these buttons under your image. These have specific actions. There are options in you right-click menu too that are quite important. In the image above, you can see U1, U2, U3, U4 - these are your upscale buttons. These move left to right from top to bottom. So the lady with big red hair bottom right is U4 if you want to upscale that one. The blue remix is to make more of the same prompt but with differing styles. In this instance, you may not get much variation.

The V1,2,3,4 are variations of the the images using that as your style. This is great for your consistency and worth trying out so that you can get a better and closer variety of poses and clothing choices. In the images below, I am making a Police Chief for my story. In the first image, I liked the one with a shorter tie to show off the round belly and the high waistline of the chief as he looks both bumbling but underhand.

Getting Consistent for Story Making

Adding your image to your Midjourney chatbot, you need to right-click on your image, and use the ‘envelope’ emoji from ‘add reaction’ if not there, to post it to your DMs.

You will notice in the prompt I have, that there are some words and letters that are preceded with -- this is a signifier for the image delivery: aspect ratio, version of generative AI within Midjourney and, here you can see the word —seed with a number. This is because I have used a link to the image I made earlier and it has a reference within MJ that is called a ‘seed’. We don’t need to worry about this. But the —ar 3:2 is the aspect ratio or the forma the image is shaped in. 16:9, for example, is HD video, 5:7 is photographs and some films, Raiders of the Lost Ark, Matrix or Forest Gump are 21:9. 1:1 is Instagram… This will help you later when you are thinking about where this is finally been seen, if at all. For a comic book, this is paneled therefore, depending on the the frame, it may be 9:16, 3:2. 4:16, 12:4 to get the feel of the story line and impact.

What I am going to do now, is make a whole bunch of characters on character sheets. Then download them. To do this, you need to run the process above, maybe experiment with - -ar 3:2 and dress your character in the clothes for your story/ journey. See the galleries below. Panel sets and characters.

Characters and scenes from 9 panel comics.

In my comic, I want a Police Chief, a Rookie Cop, a Rogue cop, a Female love interest and a Mafia boss. Then, while I have my idea of the mafia being neutral but controlled by the lady for past endevours, the rogue is in love with her, the chief and mafia are in cahoots and the lady is in love with the rookie - I add this to ChatGPT to get an idea of scenes.

Shadows of Allegiance is born

Scenes and Detail

Photoshop

To give me the consistency, I use Photoshop to cut the characters out an upload these to the chat in MJ. To do this, you need to use the lasso tool in Photoshop or Photopea at Photopea.com, and then export to a .PNG. file.

So I make new file to use and upload.

Simply click the plus button on the chat box of MJ and hit

Once you upload, you can get a link by right clicking and getting the character’s Midjourney link to train MJ and get an even more tighter variation. See here:

Giving us these close styles.

Story Settings

Of course, there are the settings that we can start to manipulate to create a story. A few Photoshop techniques of layering, blending and light matching will let you play with characters over laid with settings. Some of the pitfalls of Midjourney are the way it handles multiple people in a scene and the scene itself. It can get confused, so we still need to have a learned user of something like Photoshop to get multi-layered images to work.

And, in schools, you still need a teacher to be on hand for that human touch, that neither video nor tinkering allows.

For the middle ground in this process, I’ve used Comic Life 3 to put the paneled images together alongside the MJ images I spliced from a process before this one. This middle ground, the gives opportunity to less confirdent users to create with Craiyon and make a similar standard of comic. See here.

Video Support

For people who get this far, I’ve made a playlist of consistent characters with Midjourney.

See this content in the original post