Launch Story · Small business · 0:30

Game Builders · Game studios · 0:26

Case study · Brand launch · Leonardo AI · 2026

Built for Builders

Built for Builders

Built for Builders

Two vertical launch films and a launch email for Leonardo's visual-first API. I took it from a voice memo to the final cut: the story, the storyboard in the new brand, and the workflow that produced every asset.

2

Vertical launch films, 9:16

1

Line both films are built on: you build it once

1

Launch EDM, made from the same assets

2

Weavy flows, one per film, feeding every asset
Client

Leonardo AI

My role

Creative direction, storyboard, workflow design, EDM design, motion

Tools

Leonardo, Weavy, After Effects, Figma, ElevenLabs

Overview

Leonardo launched a visual-first API in March 2026 and needed launch content that made an API feel like something you could actually see. I was Creative Director on it. I shaped the concept, designed the storyboards in Leonardo's new brand direction, built the assets in Leonardo and Weavy, animated both films in After Effects and designed the launch EDM.

There's no big generation count in this one. I didn't track it, and honestly it isn't the interesting part. The interesting part is the system. One character template, one studio, two node flows, rolled out across two films, an email and a shelf of merch.

01 · The brief

How do you film an API?

How do you film an API?

How do you film an API?

An API is code. Nobody gets excited watching code. So the films had to land one idea anyone could see in a second: you build it once, and every customer gets their own version.

That line became the spine for both films. Each one follows a single maker, shows the moment they set something up, then pulls back to show it running for everyone else. The script is the same skeleton both times.

VO skeleton · both films

Meet Leonardo. AI for images and video. Now with a visual-first API. Built for builders. You build it once. One system. Leonardo API. Yours to grow.

For brands: if the product is invisible, show the one thing people would see change.

02 · Two films

Same structure, two very different builders

Same structure, two very different builders

Same structure, two very different builders

Splitting the launch into two audiences let each film stay simple. Same beats, same end line, different world.

Launch Story

Game Builders

Audience

Small merch businesses

Game studios and indie builders

Hero

Marco, a merch founder in his own studio, plus his scene-stealing cat

A pink-haired gamer who ends up as a player in her own game

The setup

Designs a tee graphic visually, exports ready-to-use code

Designs game assets visually, exports the code

What customers add

Their own design or photo

A selfie that becomes their character

The payoff

Every order personalised across tees, hoodies, caps and more

Every player gets a card, a skin and a place in the world

Length

30 seconds, 9:16, 24fps

26 seconds, 9:16, 24fps

Launch Story · frames

Game Builders · frames

03 · Operating model

One director, a handful of specialists

One director, a handful of specialists

One director, a handful of specialists

Director · Me

The concept, the storyboard, the asset workflow, art direction, every approval, the edit and the EDM. I also put the motion together in After Effects.

Image and video · Leonardo + Weavy

Leonardo for hero stills and product shots (Nano Banana Pro, Nano Banana 2, FLUX.2 Pro, Seedream 4.5). Weavy for node flows that chain models together, branch variations, upscale with Magnific and animate with Kling 3.

Voice · ElevenLabs, then a real person

An ElevenLabs guide VO locked the timing early so I could animate to it. The final was recorded again with a real voice, and I re-timed to the performance.

Writing partner · Claude

Cleaning up voice memos, building personas and keeping prompts short. Useful for thinking out loud, never for the final call.

None of it depends on one tool. Swap a model next month and the structure still holds.

04 · How it was made

Six things I'd do again

Six things I'd do again

Six things I'd do again

1

Start with a voice memo

The idea showed up while I was working, so I just recorded it. Rough version: a gamer puts herself into her own game, then a crowd of players shows up wearing what she made. Cleaning that up gave me the structure for both films.

Early on I swapped battle royale for co-op. Everyone using her creation together is a warmer story for an API launch than everyone fighting.

For brands: the rough first version is usually the honest one. Keep it.

2

Design a character system, not a character

Estrella started as a front and back turnaround: yellow suit, chrome jetpack with red rings, bell-shaped nozzles. Then I built a template prompt around her. Drop in anyone's photo and it keeps the suit, the jetpack and the proportions, and only swaps the personal stuff like hair, skin tone and accessories.

That template is the API idea in miniature. Once it worked, the same character could go anywhere: a pin, a plush, a hoodie, a keychain, a blind box, a phone case.

For brands: if the product scales, the creative should scale with it.

Photo in, character out, then everywhere

3

Lock the world before you shoot it

Marco started as a persona and quickly became a look: washed olive tee, an industrial loft with concrete and beams, warm window light, and a white cat who steals every scene. Once his face and studio existed as reference images, every prompt just pointed at them.

Prompts got shorter and the errors went down. Long, detailed prompts broke more often than short ones backed by good references.

For brands: your brand book should also exist as images a model can read.

Marco's studio and the tee graphics his customers personalise

4

Build it as a flow, not a folder

Weavy let me lay the pipeline out as nodes. On Game Builders, one character sheet feeds the player cards, the space station hub, the gameplay POVs, the equipment sheet, the upscales and the Kling clips. Change one input and the branches follow.

The small business board did the same for Marco's frames, the cat, the tee graphics and the product mockups.

For brands: a flow is a reusable asset. The next campaign starts halfway done.

Weavy · API Game Builder flow
Weavy · API SMB stills flow

5

Storyboard in the new brand

Leonardo had just moved to a new brand direction: violet, acid green and hot pink colour fields, warped condensed type and real product UI. I designed the storyboard in Figma in that language first, so every generated frame had a slot to land in.

The type, colour and UI were designed, never generated. That's what makes both films feel like one brand.

For brands: type, logos and UI are design work. Keep them out of the diffusion model.

6

Lock the timing with a guide voice

I generated a guide VO in ElevenLabs early, so the story had a rhythm before a single frame was final. The storyboard, the shot list and the edit all hung off that timing.

Once the cut worked, we recorded it again with a real voice and I re-timed the edit and the After Effects motion to the performance. Small thing, big difference in how it feels.

For brands: lock the voice early. Every visual decision gets easier once the timing is fixed.

The finish stays human.

AI got me to good frames fast. The story, the brand, the workflow and knowing when it was done were still the job.

05 · The launch EDM

Same system, one more surface

Same system, one more surface

Same system, one more surface

The launch email reused the film assets on purpose. Same character, same type, same colour fields, rebuilt for an inbox. I designed it in a few versions, dark and light, with a different hero for each audience.

It's the build once, use everywhere idea, applied to our own production.

Dark · scroll inside
Light
Alternate hero

06 · Guardrails and limits

What I held to, and what still needs people

What I held to, and what still needs people

What I held to, and what still needs people

Guardrails

No real game IP. I invented a co-op space game instead of borrowing a famous one. Weapons became space gear, so the world reads friendly. No third-party logos in the final frames. Every still and clip approved by me before it went into the edit. A real human voice on the final.

Where it still needs people

Models drift back to defaults. Estrella kept coming back in a blue suit until the prompt named every detail. Content filters flagged combat words around cute characters, so "firing" became "launching energy blasts". Hands, crowds and consistency across a full scene still need checking frame by frame. And taste. The model offers options. I decide.

What this means for creative teams

Build the system, then tell the story.

Build the system, then tell the story.

Build the system, then tell the story.

This was an API launch, but the lesson is about production. One character template, one studio, two flows. Set it up once and it ran across two films, a launch email and a shelf of merch. Pretty much the same idea we were selling.

AI made the stills faster. The concept, the brand, the motion and the finish were still the job, and that's the part I bring.

© 2026 Phyo Thu · @phyofox

Wellington, NZ

© 2026 Phyo Thu · @phyofox

Wellington, NZ