Christmas is approaching, and while retailers have a vast array of toys, some parents wish for a more personal gift. One option could be a custom-made picture book. However, not everyone has a knack for writing child-friendly stories or creating illustrations. Moreover, creating such a book can be time-consuming.
Thanks to generative AI, creating a picture book is now easier. ChatGPT can write stories, and AI-based image generators can create illustrations. Here, we explain the process and steps you need to follow.
We created a children’s Christmas story using ChatGPT and Midjourney. The result may not win a children’s book award, and the illustrations might need some refinement. However, it shows that you can create a picture book in a short time using AI. How much effort you put into fine-tuning is up to you.
First, you need a story for your picture book. You can write it yourself, or let ChatGPT handle it. Even when using AI, you can decide the story’s theme through your prompt. You can also include your children’s names in the story.
For our example, we used the prompt: “Write a children’s story about a boy named Joshua who has a Christmas adventure.” You can go into more detail by specifying locations, characters, or events. If you’re not satisfied with the result, you can ask ChatGPT to generate a new response. It’s also worth editing the story yourself to solve potential illustration issues, as image generators sometimes struggle with specific prompts.
To illustrate the book, you can use various AI-based image generators. We used Midjourney for our example. Regardless of the tool, you need to ensure that your character appears consistently across all images. You could train an image AI with existing images or choose a fictional character hoping for consistent results.
We chose a different approach. Midjourney allows you to include images with your text prompt. We created our main character in various poses using Midjourney. Our prompt was: “little boy character, multiple poses and expressions, children’s book illustration style, simple, cute, full color, green shoes, short blonde hair, hoodie, blue pants, flat color, –no outline.” It’s important to specify that you want the character in different poses and expressions. You can choose the character’s description, such as hair color and art style.
Once you’ve created your character, download the image with different poses and separate them into individual images using an image editing program. Then, upload them to Discord, either via drag-and-drop or the upload button. After uploading, copy the URLs of all images. Start a new image prompt with the /imagine command and paste all URLs separated by spaces, followed by the original prompt with a modification.
Replace “multiple poses and expressions” with the specific scene description. If none of the four automatically generated results work, try again. If it still doesn’t work, modify your prompt until a suitable image is created. Experimentation is key. Zooming out can help, as well as adjusting the importance of images compared to the text prompt using the –iw command. Set a value between 0 and 2, where a lower value gives more weight to the text prompt over the images. Although you don’t want to stray too far from the image templates, adjusting the value can help.
Creating an AI-generated picture book can be a fun and creative process, offering a unique and personalized gift for children during the holiday season.