惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Google DeepMind News
Google DeepMind News
博客园 - 聂微东
Microsoft Azure Blog
Microsoft Azure Blog
V
Visual Studio Blog
IT之家
IT之家
博客园 - 【当耐特】
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
B
Blog
爱范儿
爱范儿
阮一峰的网络日志
阮一峰的网络日志
云风的 BLOG
云风的 BLOG
Vercel News
Vercel News
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
H
Hackread – Cybersecurity News, Data Breaches, AI and More
H
Help Net Security
J
Java Code Geeks
aimingoo的专栏
aimingoo的专栏
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
B
Blog RSS Feed
Blog — PlanetScale
Blog — PlanetScale
S
SegmentFault 最新的问题
Apple Machine Learning Research
Apple Machine Learning Research

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
I shipped an AI cartoon generator from Timor-Leste in 2 w...
Ajito Nelson Lucio da Costa · 2026-06-18 · via DEV Community
Cover image for I shipped an AI cartoon generator from Timor-Leste in 2 weeks — here's how

Ajito Nelson Lucio da Costa

TL;DR — I built FurakToon, an AI app that
turns a sentence into a beautiful anime or cartoon image in seconds. It speaks
21 languages with a Southeast-Asia focus, has a monthly free-credit system,
and was built in ~2 weeks for the Mind the Product "Everyone Ships Now"
hackathon. This is the build-in-public story.


Why I built this

I'm building from Timor-Leste 🇹🇱. Furak means "beautiful" in Tetum, our
language — so the app is FurakToon: beautiful cartoons.

That one constraint shaped the whole product.

Home Page


What it does

  • 🖌️ Type → pick a style → generate. Anime or cartoon, 1024×1024, in seconds.
  • 🧑‍🎨 Reference faces. Upload a photo and the model redraws that person in your scene — great for "make me an anime character."
  • AI prompt enhancement. One click rewrites your rough idea into a rich prompt (Llama 3.3 70B).
  • 🛡️ Two-layer safety. Every prompt is moderated before anything is generated.
  • 🪙 Credits. 10 free per month — 1 per image, 2 when you use a reference.
  • 🖼️ Personal gallery, 🌗 dark mode by default, and 21 languages (Tetum, Indonesian, Malay, Tagalog, Thai, Vietnamese… + RTL Arabic).

Create Page


The stack

Nothing exotic — boring tech, shipped fast:

  • Next.js 16 (App Router, React 19) + Tailwind CSS v4
  • Supabase — auth, Postgres, Storage, row-level security
  • Together AI — image models (Flux, SDXL, Gemini Flash Image) + Llama for text
  • Novus.ai — product analytics (the hackathon's one required piece)
  • Built with AI-assisted development the whole way

Three things that were harder than the AI part

The image generation? A few API calls. The product around it is where the real
work was.

1. A credits system that can't be cheated

"10 free images a month" sounds trivial until you think about double-spends and
refunds. I pushed the logic into Postgres functions so spending is atomic:

-- spend_credits(): refill if a new month started, then deduct only if affordable
update public.credits
   set balance = balance - cost
 where user_id = uid and balance >= cost
returning balance into new_bal;   -- null = not enough credits

The monthly reset is lazy — no cron job. The first action in a new month
tops you back up to 10. And if generation fails after charging, the credits are
refunded automatically. New users get their 10 via a signup trigger.

Profile Page

2. 21 languages without a heavyweight library

I rolled a tiny typed i18n layer: one dictionary per locale, English as the
fallback, with {placeholder} interpolation and dir="rtl" for Arabic. The
locale persists in localStorage and resolves through useSyncExternalStore so
there's no hydration flash. Switching language is instant.

The two Timor-Leste flag pieces in the showcase strip aren't decoration —
they're the whole point.

3. The "is the database awake?" problem

I'm on Supabase's free tier, which pauses after ~7 days idle. During judging,
a sleeping DB = a dead demo. So I added:

  • a small DB-status indicator in the footer (green = connected) backed by a /api/health probe, and
  • a GitHub Action that pings the live site every few hours to keep it warm.

Small touches, but they're the difference between "demo" and "product."


What shipping with Novus taught me

The hackathon required Novus.ai, and I expected to bolt it on and forget it.
Instead it changed how I worked.

I instrumented the real moments — user_registered, image_generated,
image_downloaded, content_moderation_blocked, generation_blocked_no_credits,
language_switched — and suddenly I wasn't guessing. I could see where people
dropped off.

Two things genuinely surprised me:

  1. Novus Cortex built a sitemap of my app by itself — Public pages → Protected Create flow → API routes → Gallery. It understood the architecture from usage, not from my code.
  2. Novus opened pull requests on my repo. Real, reviewable changes I could merge — AI feedback turned into AI contribution.

That's the shift: from "I think this is good" to "I can see what's working,
and act on it."

Novus Dashboard

-

Novus Site Map


Lessons for anyone shipping fast

  • Pick one sharp constraint. "Must work in SE Asian languages" made every decision easier.
  • Put the hard logic in the database. Atomic credit spending in Postgres beat any client-side guard.
  • Instrument from day one. Analytics isn't a launch chore — it's how you steer while building.
  • Sweat the "is it actually live" details. Keep the DB awake, test in incognito, default to your best-looking theme.

Try it

👉 Live app: furaktoon.fun
👉 Code: github.com/ajitonelsonn/FurakToon

Type an idea, pick anime or cartoon, and make something beautiful — in your
language.

Built for World Product Day 2026 — "Everyone Ships Now" by Mind the Product
× Novus.ai. Furak means "beautiful" in Tetum 🇹🇱

EveryoneShipsNow