AI design tool, a bit ahead of its time

Musho

2023Product lead · Design · Concept · Creative directionMusho App (Figma)

Type it. Get a design.

Musho turned a prompt into a real, editable layout… inside Figma, before Figma had any AI at all.

The idea was simple to say and hard to build: write what you want, and Musho composes the whole thing on the canvas. Copy, imagery, color palettes, layout. Not a picture of a design… actual layers you could grab and move.

We shipped it as a Figma plugin in 2023. This was before the big models were everywhere, before Figma had a single AI button. So a lot of people looked at it and went huh, and a lot of people looked at it and went whoa. Both reactions felt right.

Prompt in, design out

Prompt in, design out

On the canvas

The magic trick was that it landed as editable Figma layers, not a flat export. You could ask for a landing page for a puppy spa, get real headlines, real sections, real color, and then just keep working. Change a word, swap the palette, drag a block.

That was the whole bet. An AI that hands you a starting point you actually want to touch, instead of one you have to redo.

Watch it compose

You typed a prompt, Musho thought for a second, and a layout assembled itself piece by piece. Copy first, then imagery, then the palette settling into place. Every run felt a little like a slot machine you were rooting for.

Generation in action
Generation in action
Prompt to layout
Prompt to layout
A generated page, selection handles still on
A generated page, selection handles still on
Editable, always
Editable, always
Sections you could reshuffle
Sections you could reshuffle
Real Figma layers underneath
Real Figma layers underneath

Palettes and pieces

Color that had taste

Prompts are easy to answer badly. The hard part was making the output feel considered. So we spent real time on the palette engine, the type pairings, the spacing, the little defaults that separate a design from a mess of boxes.

Folios reel. Whole layouts from a line of text.

It grew

A plugin became a small ecosystem. A website so you didn't have to open Figma to try it. Template generation tuned to a brand. A curated AI image library so the pictures weren't garbage. And a sub-brand, Aisplash, for the imagery side of things.

We built it as a team. I led product, design, and the concept, which mostly meant arguing about what good output even was, then drawing it.

The website, no Figma required
Aisplash, the image side
The Musho mark
The Musho mark
The brand board, waiting to be filled
The brand board, waiting to be filled
Aisplash sub-brand
Aisplash sub-brand

Out in the world

Real sites from a prompt

The demos that landed were the specific ones. A puppy spa. A UX agency. You typed the brief a normal client would give you, and Musho built the page. People got it fastest when the example was something they could picture selling.

Showing it in person
Showing it in person
The huh-to-whoa pipeline, live
The huh-to-whoa pipeline, live
Real rooms, real reactions
Real rooms, real reactions

Keep digging

Type a prompt, hit generate

musho

product

Innovative UX, take two

musho

product
UI meets AI, full yellow

UI meets AI, full yellow

musho

brand

Puppy spa, full video

musho

product

Main color, accent color

musho

product

Musho was right about the shape of things and early on the timing. A year later everyone had an AI design button. That's the deal with being a bit ahead… you get to be right and tired at the same time. Un abrazote to the whole team.

Next up

Robyn

The animated series we almost made with TIME Studios.