What We Learned

The episodes themselves are fun. Silly. Weird. No one will ever call this high art, but we’re proud of them. Not just because of the hours of late-night eruptive laughter they brought us while we created them, but also because of what we learned along the way. Because what started as a funny AI experiment to kill time on a plane turned into a masterclass in what it actually means to tell stories with these new tools at our disposal. Here’s four brief lessons we learned on storytelling in the age of AI:

Lesson One: AI Can Organize, But It Can’t Originate

AI is brilliant at structure. It can lay out scenes, break down a script into beats, even spit out shot lists. But the soul of a story? That has to come from humans.

Our tiny selves shouldn’t just say funny things—they should have big personalities. Style. Rhythm. Sarcasm. They should bicker and joke around like we do. They should think and behave like exaggerated versions of us. And no prompt can do that alone. The characters work because they’re ours. Our memories. Our dynamic. Our sense of timing and delivery.

We used AI to help build the skeleton—but the muscles, the voice, the soul? That’s 100% human. Still is.

Lesson Two: Real Voice Carries Real Weight

We’ve worked with AI-powered voice applications before. Years ago. Back when it was still called machine learning. We know how powerful synthetic voice has become.

But for this project, we chose to use real voice recordings. Because even with the best AI voice tools today, you lose the emotional flexibility that a human can produce instantly. When we say something with sadness, tension, or joy—it doesn’t require computing power, tokens or rendering. It just is. That rawness and realness matters.

We’re not against synthetic voice. For many use cases, it’s amazing. But when you're trying to evoke feeling, we found that nothing matches the tone, timing, or emotional fidelity of a human voice—flaws and all.

Lesson Three: AI Video Is Getting There—but Not Quite

Making a 2-minute AI video sounds easy. It’s not.

Each Tiny Tales episode took 25 to 35 shots. Not takes—shots. Each had to be thought through and painstakingly directed. What expression do we want here? What are the props? How should the milk move? How do we add a hat, pull up a hood or clutch a microphone without breaking the character’s continuity?

Even the best tools today struggle with consistency, especially across multiple scenes. Characters shift. Backgrounds change. Costumes morph unexpectedly. A plump little baby hand becomes a scary alien claw. Getting a series of shots to feel like one scene takes iteration, editing, a lot of workaround creativity, and at least one glass of decent red wine!

We had to become set designers, sound editors, and digital babysitters—all while experimenting with multiple platforms to stitch it together. Lemon. Kling. Hedra. Eleven Labs. And the list goes on.

Lesson Four: Assembly Is the New Art

Once we had the story, the voices, and the video, the real challenge began: making it all work. Putting Humpty Dumpty back together again, if you will. That’s the part nobody talks about.

Editing. Color grading. Sound design. Syncing. Audio leveling. Music. Flow.

This is where the storyteller shows up and shows out. The AI tools gave us fragments—but it was still our job to decide how to piece them together. Where to cut. Where to linger. Where to surprise. You can’t automate taste.

We’re entering a phase where the barrier to creation is lower than ever—but the need for craft is just as high. Knowing how and when to use the tools is the new superpower.

We’re happy to share the first episode of Tiny Tales today. We’ll be sharing other episodes shortly. It’s not perfect. But it’s real. And it’s ours.

We hope it makes you laugh—and learn and think.

Because this is just the beginning. Not just for us, but for what AI and storytelling can do together, when the people behind the tools bring their whole selves to the work.

– Fabio & Ross