ss5
ss4
ss7
ss3
ss2
ss6
ss1
Logo
Feature graphic

The spark

I’ve always loved manga — the pacing of a good panel, the quiet beat before a twist, the way a single speech bubble can carry a whole chapter. For years I sketched ideas in notes apps and never finished them. Drawing every frame takes forever. Writing alone felt lonely.

I wanted something that lowered that first blank-page fear: pick a cast, write a prompt, get inked pages you can actually read, continue, and share. Not a toy that dumps one random image — a tiny studio for people who love stories and want to start telling them.

That’s HAKI: turn heroes + a story prompt into multi-panel manga, manhwa, manhua, or comics — then keep going.

Inspiration

The name nods to haki — presence, will, the energy that makes a character feel real. The product idea is simpler:

Everyone has a story. Most people never get past page one.

HAKI is built so someone who loves manga can:

Cast heroes (from a photo, AI design, or as-is)
Write a scene (or tap Inspire me)
Generate a 3- or 6-panel page in the style they care about
Edit lettering and layout, continue the chapter, add a twist
Share a page, PDF, or a 9:16 reel

The goal isn’t “AI replaces artists.” It’s start writing and sharing — the same impulse that made me fall for manga in the first place.

What I built

HAKI is a Flutter app (iOS + Android) with:

Create — cast, visual style (Manga / Comic / Manhwa / Manhua), audience tone (Shonen, Shojo, Seinen, Josei…), story prompt
Library — series with chapters, pages, and a cast bible for continuity
Hero studio — design or stylize characters once, reuse them across stories
Reader + editor — read RTL/LTR or long-strip; tweak panels, regenerate, continue
Export — share images, save to gallery, PDF, vertical reel for social

Monetization runs through RevenueCat: Pro / credits paywalls, free trial path, and credit gates on avatar + panel generation so shipping and sustainability stay honest.

Rough credit model for a first page (panels only):

$$ C_{\text{page}} = n_{\text{panels}} \times c_{\text{panel}} \quad\text{with}\quad n_{\text{panels}} \in {3,6},\quad c_{\text{panel}} = 10 $$

so a first manga page costs (C = 30) or (C = 60) credits — clear in the CTA before you tap Create.

Under the hood: Firebase callables for storyboard → panel art → lettering, a cast bible so characters stay consistent, and client-side composition for layouts, export, and reels.

What I learned

Continuity matters more than a pretty single panel. A manga people want to share needs the same faces across pages. Casting heroes first changed everything.
Style ≠ tone. Visual format (manga ink vs color manhwa) and audience tone (shonen vs shojo) are separate axes. Mixing them made prompts better and users happier.
Shipping monetization early saves pain. RevenueCat + credit pre-checks on the client and deduction on the server kept “fun AI” from becoming an unbounded bill.
Share is part of the product. If the loop stops at “nice preview,” people don’t come back. Page / chapter / series export and reels close the loop: write → generate → share.

How I built it

Core loop first — cast + prompt → storyboard → panels → preview
Make it a series — chapters, continue / twist, cast discoveries mid-story
Make it yours — editor, reader prefs per style, library drafts
Make it shippable — RevenueCat paywalls, quotas, AI consent, content reporting
Make it shareable — compose, PDF, 9:16 reel, branded footer

Design was dark “comic deck” UI; copy and onboarding push one idea: be the hero of your story.

Challenges

Consistency. Early generations looked cool once, then the hero mutated. Fixing that meant a real cast bible in prompts and locking focal characters — harder than “one big image prompt.”

Latency & cost. A page is many model calls (script + (n) images + lettering). We charged per panel image, not for scripting, and showed cost up front so users aren’t surprised:

$$ T_{\text{page}} \approx T_{\text{script}} + n \cdot T_{\text{image}} + T_{\text{letter}} $$

Layout vs lettering. Bubbles and SFX have to land in real panel slots. Vision lettering + client layout mapping was fiddly; silent fallbacks when the script soft-fails kept the app usable.

Policy & trust. Photo → character needs clear consent; AI images need in-app reporting. That’s not glamorous, but it’s required to ship and keep the community safe.

Scope under Shipaton. Easy to drown in features. We cut a “Plan” tab and stayed brutal: Create, Library, generate, continue, share.

Why it matters

Manga culture is full of people who want to write — they just never get ink on the page. HAKI is for them: start a story tonight, star as the lead if you want, keep the cast consistent, and share a page or a reel before the idea fades.

If Shipaton is about shipping something people can use and pay for, this is ours: a small studio for big stories — powered by AI, sustained by RevenueCat, built for anyone who’s ever said “I should turn this into a manga.”