Using GPT 2 to make a comic book page

Table of Contents Creating a comic book page with a basic prompt Creating a comic book page with a start image of your characters Generating a comic book page with two start images, and designating specific actions for the last few panels Generating a comic book page with a written story with dialogue, no start image Generating a comic book page with a prewritten story narrative and no dialogue, several start images Conclusion You can use GPT 2 to create a comic book page. I usually use the 16:9 size as it can fit several panels on it quite easily. You can use GPT 2 Low at 30 credits but I usually do GPT 2 Medium at 60 credits especially if there is more text or the prompt is a bit elaborate. The Medium seems to handle making a comic page better and with the less spelling errors, I find I don't have to regenerate as many time to get it right. Creating a comic book page with a basic prompt If you don't have any lines or an image, you can suggest GPT create a comic book by prompting in text to image. Like I did with this prompt: Make a comic book page about a black elegant Parisian cat showing off the sights to an Australian koala tourist The important bits you need. The instruction: "Make a comic book page about ... " You can also do "Make a graphic novel page about ... " or a "Sunday newspaper cartoon page", etc. They will have a different vibe. This is the framework for your page. The actors/subject: You need the main actors or the subject in the prompt. I did a black cat and a koala. Adjectives and descriptions give the AI clues to their character and will effect their story. Action: What you want to happen in the page. I have the cat "showing the sights" Location/Era : I said my cat was Parisian so the AI assumed they were in Paris. If I wanted a specific place or time period, it needs to be in the prompt and will affect the story. It will generate the story for you with text. As sometimes the words may be mispelled, you may have to generate the same prompt a few times until you get perfect spelling. If you have a graphic program, you can correct spelling errors that way instead of making a new comic book page. viola! The page! Creating a comic book page with a start image If you have an image with the comic book characters, then use image-to-image to create the comic book. Pay attention to the style of your image because sometimes it will affect your cartoon page. If you want the style, then say so in the prompt. So here I started with this page. Then I used this prompt: Using the black cat and the koala and the art style in image 1, make a comic book page about the koala showing off the sights to the cat tourist The important bits you need. Identify what it is in image 1 you want to use: In this case, I wanted to use the cat, the koala and the art style so I started the prompt with "Using the black cat and the koala and the art style in image 1" The instruction : Then I added, "Make a comic book page about ... " As I said before if you prompt it for a "graphic novel" or "Sunday newspaper comics", these are framework for your page. If you don't want the Image style to be your comic book style because you only want the characters, you will need to define your comic book style. The action/story : give a brief description of what is happening. In this case "the koala showing off the sights to the cat tourist". Viola! Here's the page Now if I had only wanted the cat and the koala but not the style, my prompt would not mention the style: Using the black cat and the koala in image 1, make a comic book page about the koala showing off the sights to the cat tourist And the AI will generate what it pleases like it did here: Generating a comic book page with two start images, and designating specific actions for the last few panels Sometimes you will have your character whom you already have an image of and an idea for different actions in different panels. Once again we will start with image-to-image In this case I had 2 start images: the wizard and the sky-kraken In the image above you can see I have the kraken fighting the ship as image 1 and the wizard as image 2. This is my prompt: Create a graphic novel page with a few panels. Show the wizard in Image 2 on the airship and battling the sky-kraken in image 1. The wizard's staff in his right hand shoots out electricity. In the wizard's left hand he carries an opened bottle of green-glowing radioactive vodka. The wizard from image 2 faces the kraken and points his staff shooting a lightning spell at the kraken. The third to last panel shows the wizard spilling the vodka on a nearby tentacle. The second to last panel shows the wizard shooting a shrinking spell from hist staff at the kraken. The last panel shows the kraken shrinking to the size of a mouse. No speckled noise, no random particles, no visual artifacts. This one is a bit more intricate than previously. The instruction : I started with the instruction "Create a graphic novel page with a few panels. " You can just say graphic novel page. You can say a few panels or 6 panels. It's up to you. Identify what you want to use in the images : I wrote "Show the wizard in Image 2 on the airship and battling the sky-kraken in image 1." So right away, the AI is told I wanted the wizard in Image 2. From image 1, I want the airship and the sky-kraken. Describe anything you want to be different from the image: I wanted the wizard's staff to show it's magic and I wanted the wizard to be holding green vodka so I wrote "The wizard's staff in his right hand shoots out electricity. In the wizard's left hand he carries an opened bottle of green-glowing radioactive vodka. " If you know what scenes you want to appear in each panel describe them. You can specify which panel you want a specific action to appear. The next few sentences are seperated to show you what I wanted in each panel. I knew how I wanted the the story to end so while I didn't specify the first few panels (letting the AI decide), I did specify the last 3 panels The wizard from image 2 faces the kraken and points his staff shooting a lightning spell at the kraken. The third to last panel shows the wizard spilling the vodka on a nearby tentacle. The second to last panel shows the wizard shooting a shrinking spell from hist staff at the kraken. The last panel shows the kraken shrinking to the size of a mouse. Dialogue/narrative: I did not specify any dialogue or narration so the AI made it up. Check the spelling as you might have to regenerate or fix spelling in an outside graphics program. It does spell pretty well most of the time. The added dialogue can be fun. You can add your own dialogue if you prefer. Sometimes you don't want any added dialogue or text, just the images. and you can add "no text" to your prompt. Generating a comic book page with a written story with dialogue, no start image Sometimes you have the story written and the dialogue. Basically you just drop your story in with a bit of description of your characters. So this was the prompt: Make a comic book page with two women (a boho new-age woman is named Rie and the casual Asian American woman is named Cat) and this bit of story: Rie smiling: "That's a perfect idea. Let's go. I know a great magical coffee shop." Cat looks at Rie. Cat: "Are you high?" Rie: "Come on. Besides they really do have great coffee." Cat thinks, 'Honestly Rie gets a little woo-woo sometimes but she does know coffee. If she says the coffee is great, it's magnificient.' Cat: "Ok, I need some caffeine anyways." Cat and Rie head down some steps to a basement coffee shop. When they open the door, Cat: "What is this place???" Rie:"It's a coffee shop." Cat: "But that's a dragon?!!" Rie: "Oh, yeah, Henry. He's cool. You might ask him. He might know a wizard. Don't ask the fairy tho. You have no training in negotiating with a fairy and you'll end up with a bad deal." So the important bits "Make a comic book" Who or what : two women described as boho and casual identify where the prompt ends and the story starts: "this bit of story:" dialogue : be very clear who is speaking. I found making it look like a movie script helps with clarifying the speaker. Generating a comic book page with a prewritten story narrative and no dialogue, several start images Ok so this time we're starting with 4 start images. Image 1 and 2 are reference images for the main character. third image is the reference image for the fairy; 4th image references the wizard the prompt: Create a graphic novel page. No speckled noise, no random particles, no visual artifacts. The narrator of the story is the asian american woman in image 1 and 2. the fairy the narrator sits down to talk to in the coffee shop is the fairy in image 3. When the narrator ends up in another world, she is sitting in a magical cafe with the wizard in image 4 and an anthropomorphic dragon. A human-sized fairy works at the counter serving coffee. The first panel should be the narrator talking with the fairy. Here is the narration: "But of course I talked to the fairy because the fairy looked nice and I didn't think I'm up to talking with a dragon. I didn't even believe in them a day ago. Well I didn't believe in fairies either but she looks almost like a real human person, you know? And that's how I ended up here. I'm not on earth. I don't know where I am. I don't even know what year it is. And now I'm really looking for a wizard so I can get back home to Seattle 2026. Well, maybe I'll travel a bit first, see the sights. I've never been out of Seattle before. First though maybe I need to get a job. I still have my denim duster, my clothes, and my bag from home. Nothing else. But I am NOT going to ask this fairy for a job. The last time I asked a fairy for help ... well, here we are. Somewhere far far away ...." So the important bits "Make a graphic novel page" Image 1&2 : The narrator of the story is the asian american woman in image 1 and 2 Image 3 : the fairy the narrator sits down to talk to in the coffee shop is the fairy in image 3 Image 4 : the wizard in image 4 identify where the prompt ends and the story starts: "Here is the narration:" I initially said "Here is the story: ...". Saying "story" was a mistake here because it was all narration so I changed it to "narration". With a story, characters might dialogue. With narration, in a comic book, you often have little text boxes which is what I wanted here. Conclusion I tried to show you how to do all the scenarios I could think of that you might need in creating our comic book, graphic novel, cartoon strip or illustrated book whether you want the AI to generate the story and characters or you have a story complete with dialogue written out and characters you have already created. If you have a question that I didn't cover about creating a comic book/graphic novel page, just drop the question in the comments. Have fun! Let me know what you make.