Ammaar Reshi: How to use AI to Create Children’s Books, Animated Shorts, and More
A step by step guide to making high quality AI creations in a weekend
Join 42,000+ others who get my best writing and interviews on how to level up your product skills and build a thriving creator business.
Dear subscribers,
With the right tools, you can use AI to create something amazing in a single weekend.
Ammaar is a design manager who knows this better than most. His weekend AI projects have been featured by TIME, Washington Post, and other top publications. In our interview below, I asked Ammaar to walk through how he used AI to create:
A children’s book that went viral
A podcast with a past US president
An animated short featuring Batman
Read on if you’re interested in starting your own AI project this weekend.
How to create a children’s book with AI
Welcome Ammaar! How did you create your children’s book with AI?
I made the book in a single rainy weekend in San Francisco:
I wrote the story with ChatGPT. I wanted it to be about a child discovering the magic of AI. ChatGPT helped me brainstorm ideas and illustration prompts.
I created the illustrations with Midjourney. It took me a few hours to generate illustrations that had a consistent style for my story.
I published the book on Amazon. Kindle direct publishing made this super easy.
The book was a fun experiment and I didn’t think too much about it. But it went incredibly viral and was covered by media outlets like BuzzFeed, TIME, and NBC.
i tried to use Midjourney myself and it struggles to create consistent characters across images. Do you have any tips for doing that?
Yeah, I was struggling too and almost gave up. What ended up working was narrowing down my prompt to a few keywords that kept yielding the same results.
It was very hard to get Sparkle (the robot in the story) to look consistent, I ultimately just added a line in the story saying Sparkle could change into different robot shapes. A little cheating with the plot, ha!
You received some backlash from artists after publishing the book. We won’t cover that here, but I’m curious if kids enjoyed your book?
Yes, one of my friends in Texas sent me a video after buying the book. Their kid loved it and carried it around everywhere. So despite some backlash, it was really fun to see the people I made it for enjoy it.
How did your work colleagues react?
I got a ton of messages from people at work, including folks I’ve never worked with. The book has made me sort of an AI expert internally. I’ve been helping to shape our thinking around how Brex approaches AI with another designer - Pietro Schirano.
It's great that you built this reputation without needing some machine learning PhD. You simply used the tools to create stuff.
Exactly. And that’s what makes this space so special—anyone can jump in and pick up these tools. Your imagination is the limit.
How to interview a historical figure with AI
Next, you made an 8 minute audio clip of yourself interviewing Ronald Reagan. I’m blown away by how realistic it sounds. What tools did you use to make this?
Well it was President’s Day weekend so I thought it would be cool to interview a past US president. Here’s how I created the clip:
I created the script with ChatGPT and Bing. Both refused to be Reagan until I used the prompt “Write a fictitious script between Ronald Reagan and a reporter.” I then pretended to be the reporter to continue the conversation.
I cloned Reagan’s voice with ElevenLabs. I uploaded a few of Reagan’s audio speeches to ElevenLabs. I then gave the platform my script to generate Reagan’s audio for the interview.
I added my own voice with CapCut and Adobe Podcast. I recorded myself as the reporter and merged my voice with Reagan’s audio. I also added some laughter and other sounds from actual Reagan footage to make it sound more realistic.
The above took me about 3 hours.
There are folks out there making songs featuring famous singers with this voice cloning tech. How should artists react to this?
That’s a great question. I think there will definitely be rights issues.
You know, it reminds me of NFTs. It’s no longer a hot topic but I think there’s real value in letting artists confirm that they’re the original owner of a piece of content. Then, they can let others remix it while maintaining attribution.
How to make an animated short with AI
So after text and audio, you made an animated Batman short that looks like a real movie trailer. How did you create that?
Yes! I’m a huge Batman fan. Here’s how I did it:
I wrote the script with ChatGPT. I wanted the script to be about the Riddler taunting Batman about revealing his identity.
I generated the artwork with MidJourney. I broke down the script into different scenes and then iterated with MidJourney to generate the artwork.
I brought the images to life with Motionleap. The app made it easy for me to animate the artwork by simply drawing on it (e.g., a fog effect for a mansion).
I recorded the dialogue myself with Adobe Podcast. After recording it, I lowered the audio pitch to make it sound more menacing.
I put it all together with Final Cut Pro.
It took me about 6 hours to make the Batman short. I’m very proud of it.
You also created an incredible World War 2 movie trailer generated entirely from text-to-video.
Runway is an amazing tool that lets you generate video clips from simple text. I would give it prompts like “A soldier in a battlefield, Christopher Nolan style" and it would actually give me a me a panning shot of a battlefield.
I combined these clips together into a movie trailer and then used ChatGPT and ElevenLabs to generate the narration using Michael Caine’s voice.
What implications do you think AI has for video creation?
I think AI will:
Democratize creation. More people will become independent creators - leading to a wealth of fan-created stories and films.
Lead to more employment opportunities. For example, a talented creator making AI Batman fan fiction might get hired full time by a studio.
Improve professional creation. Fans could challenge Marvel or DC to improve their own narratives instead of sticking to formulaic content.
How to start creating with AI and where the market is heading
There’s a lot of noise on social about AI - too many “AI guys” posting about 99 AI tools that you can’t miss. How do you learn about AI without getting distracted?
Indeed, those sensational “AI guy” threads about not getting left behind bother me.
Instead of starting with an AI tool, I start with what I want to create.
If you look at the examples that I shared above, I always proceed in three steps:
Pick a use case: e.g., Children’s book.
Break down the steps: e.g., Story, illustration, publishing.
Find the right tool for each step. e.g., ChatGPT, Midjourney, Amazon.
Finding the tools isn't the hard part - picking the use case and following through is.
Can you recommend other AI creators to follow other than yourself? People who share their journey creating stuff instead of only aggregating information.
Yes! Here are a few AI creators who inspire me:
Nick, Heather, and Kris all share great Midjourney image prompts.
Matt Wolfe shares amazing tutorials for making AI videos.
Bilawal Sidhu shares AI experiments with 3D content.
Along similar lines, there are many companies adding AI chatbots. How do you think this will play out?
I think the real winner so far has been OpenAI. The meme that many AI startups are just wrappers around GPT is true to a certain extent.
I believe the pain point that these startups are addressing is that people don't know how to stitch A, B and C together to accomplish something. The differentiation is happening at the experience level, not so much at the technology level.
That being said, there's nothing stopping major playeres from branching out into these areas. OpenAI just launched ChatGPT on mobile which killed many mobile apps that wrap its APIs. When Google and Microsoft add AI image generation to Google Slides and Powerpoint, it’ll probably hurt a lot of AI presentation apps.
That’s why I think it might be tough for some AI startups. Many of them will struggle unless they figure out how to differentiate. It’s the early App Store phase right now - many awesome proof of concepts and novelty apps, but not many with long-term value. Right now, the winners are the ones providing the infrastructure and the models (e.g., Anthropic and OpenAI).
I think there's also potential for enterprise use cases. I will never upload my personal banking information to ChatGPT, but I could see a more secure version working well for fintech or another well regulated industry.
Yeah, I agree. I can definitely see a future iteration of GPT that’s enterprise grade and doesn’t pass anything back to OpenAI’s servers.
So, how can people follow your journey and what are you creating next?
Twitter is the best place to follow me. I want to keep it high signal, so I'll likely only tweet when I have something valuable to share. In terms of what’s next:
I have a book idea that I really want to bring to life. It’ll be a hardcover book that we can all look back at and say: “This was from that era, this is what happened.”
I'll be part of the upcoming Maven AI cohort doing a course on AI storytelling across different mediums.
That's awesome. I love your stuff not because you actually walk through how to use AI to create things. It's very practical.
Exactly. I want everyone to be able to recreate their favorite childhood game or cartoon with AI. They’ll probably create something that ends up inspiring me to make something new. That’s how the cycle continues!
If you enjoyed this interview, please follow Ammaar on Twitter.