I can still remember my mom telling me stories about Marvin the Magician when I was little, even if the details are fuzzy now. With my twins turning 1 this weekend, my mom flew in to celebrate with us. Tonight, while she was telling my eldest daughter a story, I got this idea: why not use AI to turn it into a real kids' book? So, I started playing around with some AI tools, using my mom's story as a starting point. Here's what I came up with:
How I Did ItÂ
To start, I record a 5 min audio clip of my Mom telling my daughter a story which I then transcribed.
Step 1
I fed this totally unstructured recording into AI Studio using Gemini 1.5 with the following prompt:
You are an experienced children's author, the skills of which are comparable to Shel Silverstein and Beatrix Potter. You were given this raw output from a grandmother telling a story to her granddaughter. Turn it into a well formed children's story perfect for a 3 year old: [Story File]
It output a rather well structured story, but it was a single block of text.
Step 2
I then decided to ask the model to help me break the story into sections where images could accompany them. I also used Gemini to help me generate prompts for the images; I used this prompt:
Rewrite this story but break it into sections that would be natural page breaks. For each page I want to generate an image. To help me do this, for each page break you suggest can you also write a prompt that I can use to generate a great image for that section.Â
Use the following format for the image prompts:
Image 1: <prompt>
Image 2: <prompt>
...
where <prompt> is the suggested prompt to generate a corresponding image to that section.
Include these image prompts in-between each page break section you have identified.
Step 3
The following were the outputs from the image generation prompts:
Image 1: A little girl with brown curly hair wearing pink rain boots and a yellow raincoat looks sadly out her window at the pouring rain.
Image 2: A little girl stands in the middle of a grand castle hall with a surprised look on her face. Her diverse group of friends are running away to hide behind pillars, furniture and large doorways.
Image 3: A friendly cartoon magician with a blue pointy hat and a long white beard presents a sparkly pink wand to a little girl with brown curly hair.
Image 4: A long table with a pink and blue tablecloth is laden with treats. There's a large bowl of punch, a big yellow birthday cake and plates of hotdogs. Five children with party hats are gathered around the table with excited expressions.
Image 5: The five children from before are gathered around a little girl with a party hat, singing happy birthday.
I then used Adobe Firefly to help me generate each image.
Putting it All Together
I took the images from Firefly, put each into a Google Slide along with the corresponding section of my story, and saved each as an image.
Now obviously there are about a thousand ways I could improve this - and I’m excited to iterate further on this…but for a 45-minute exercise, I’m pretty impressed with the final result!