I’m a developer, not a designer. The UI for Soulfire reflected that perfectly for way too long. It was the app design equivalent of what game developers call “developer art”. Functional, but not pretty.

Soulfire is an AI-powered interactive fiction app. Somewhere between text adventure game and AI roleplay. Visit virtual worlds, make choices, roleplay with AI characters, and build immersive stories. The experience is rich. But the interface? Functional. It worked, but it was holding the whole experience back.

The gap between function and feeling

Here’s the thing about interactive fiction: immersion is everything. The UI is the threshold between reality and fiction. If your UI isn’t engaging, using the app starts to feel like work.

My UI was a list of buttons. It got you where you needed to go, but it didn’t set a mood. It said “here are your options, pick one” when it should’ve been whispering “adventure awaits.”

The original Soulfire home screen

The original home screen. Functional? Yes. Inspiring? Not exactly.

Chat was standard bubbles on a gradient. The play view was plain text on a dark screen. Nothing was bad, but it didn’t exactly excite the user.

Google Stitch gave me the vision I couldn’t find

I’d been putting off a redesign because I didn’t know what “good” looked like. I knew the UI wasn’t right, but I couldn’t picture what should replace it. I have used Figma before but I dislike Figma with a passion. It never made any sense to me. Anything someone builds in Figma needs to be rebuilt from scratch elsewhere. Why not just do the second part? Figma without HTML/CSS export is like a cafe that sells pictures of food.

So yesterday I tried Google Stitch.

I showed it a few screenshots from Soulfire and typed a brief description of the app. Stitch then generated full mockups for mobile and desktop. I was blown away immediately. I was looking at the future of Soulfire’s UI, right there.

Stitch worlds dashboard Stitch character list Stitch play view mockup

Google Stitch mockups. World cards with atmospheric images, character lists with avatars. Play view with full background art. This was the vision I wanted.

It nailed things I hadn’t even articulated. I had wanted AI generated images from the start but had been putting it off because it required more infra. But seeing it there in the mockup was all I needed. A play view where background art bleeds through so the story feels like it’s inside the world. Character portraits that give each AI personality a face. I was sold.

These weren’t rough wireframes. They were polished and specific. Stitch gave me the actual UI, in both desktop and mobile views. I could prompt further and tweak it, and I did.

From mockup to code with Codex

One thing I was excited to try: Stitch doesn’t just give you images. It also generates a project markdown document describing components, layouts, and interactions. Basically a buildable spec.

I took the mockups and that spec, and fed them to OpenAI Codex. Soulfire’s frontend is Rust/Dioxus. Codex worked through it piece by piece: new layouts, card-based world browser, atmospheric play view, reworked navigation. I reviewed each task, nudged it when it drifted, and kept things moving. GPT-5.4 is amazing.

I shipped the first update that evening. Then half-way through the next day, I completed the redesign. 1.5 days from typing some ideas into Stitch, to having the completed redesign deployed in production.

Soulfire is not a toy app. The backend is 19000 lines of Rust code. The frontend, 14000 lines of Rust code. I’ve been building it for over a year. These changes touched both frontend and backend. In addition to the obvious UI changes, I also needed new paginated search endpoints, and database model changes for the new images, along with object storage support.

The redesign

New home screen New play view New chat

The new mobile UI. Home, play view, and chat. The story doesn't feel displayed anymore. It feels like you're inside it.

I had already started reworking the layout of the desktop view previously. But this new design took it much further. Wider layouts, sidebar navigation, same design language carrying across screen sizes.

Desktop worlds browser Desktop play view

Desktop. Same soul, more room to breathe.

The real bottleneck was never code

The bottleneck wasn’t implementation. It was vision. I could’ve built any of this months ago, but I couldn’t see what to build.

The new design seems obvious in hindsight, but seeing it in mockups from Stitch is what took me from, “I should make it look better someday, somehow” to “I need to create that”.

I’m not getting paid to write this. I just genuinely liked what Stitch built for me. For me it’s a game-changer and I can’t wait to throw my other apps at it and see what it comes up with.

Try the redesigned Soulfire at soulfire.app.