The problem MyMiniEpic solves

Every parent and pet owner wants a gift that feels one-of-a-kind — but personalized books today are expensive, slow, and require design skills.

Walk into any bookstore and you'll find thousands of children's books. None of them feature your child. None of them star your dog. Generic stories get read once and forgotten. But a story where your 4-year-old daughter defeats a dragon using her actual personality traits, or where your golden retriever named Biscuit saves the neighborhood — that gets read a hundred times and kept forever.

The problem is that custom illustrated books have historically been a luxury. Traditional personalized book services charge $60–$120, take 2–4 weeks to deliver, and still use rigid templates where you're just swapping a name into pre-written text. That's not personalization — that's a mail merge.

MyMiniEpic fixes this entirely.

A customer takes a short quiz — name, personality traits, favorite genre, breed (for pets) — and our AI generates a fully unique 10-page illustrated storybook starring their child or pet as the hero. Not template-swapped. Actually unique — the plot, the dialogue, the humor, the illustrations are all generated around that specific character.

What people use it for:

• A birthday gift for a child that no one else on earth has — their name, their personality, their adventure

• A keepsake for pet owners who want to immortalize their dog or cat in a real, printed storybook

• A thoughtful, personal gift for baby showers, holidays, or "just because" — the kind of gift that makes someone say "how did you even find this?"

• A memorial book for a pet that has passed — turning memories into a story that lasts

Three tiers, every budget covered:

• Digital PDF ($14.99) — instant download, readable on any device, shareable with family

• Softcover print ($29.99) — a real book, shipped to your door

• Hardcover print ($49.99) — premium quality, built to survive years of bedtime readings

**The charity angle: **Every purchase donates $1–$3 to animal shelters. Customers aren't just buying a book — they're helping a real animal while celebrating their own. We surface this as a surprise charity receipt included with the order, turning a purchase into a feel-good moment.

Why this matters beyond gifts: We've observed that stories create stronger emotional memories than photographs. A photo captures a moment. A story captures a personality. Parents and pet owners don't just want to remember what their child or pet looked like — they want to remember who they were. MyMiniEpic turns that instinct into a product.

The marketplace layer: Beyond direct-to-consumer sales, MyMiniEpic actively onboards pet shops, gift stores, and local boutiques as resellers on our platform. These shops(I have included Etsy and Shopify) can offer personalized gifts and items for the loved ones. This turns every local pet shop into a distribution channel. For these distribution channels, we find leads in CRM section.

Marketing through Google Ads Layer: We are using Google Ads to sponsor our pages and the platform we are running to reach to audience.

Challenges we ran into

Port mismatch between Vite and LocusBuild

The Vite dev server defaults to port 5173, but the .locusbuild configuration expected port 3000. The build verifier checks the port declared in .locusbuild, so even a perfectly working server on 5173 would fail verification every time — the health check would hit port 3000 and get nothing. This was frustrating because the app worked fine locally but kept failing in the Locus pipeline. The fix was hardcoding port: 3000 in both vite.config.ts and .locusbuild simultaneously so both systems agreed on where the server lives.

Backend process silently exiting during build verification

The build verifier kept reporting that the backend (tsx src/server.ts) would start and then immediately exit, failing the health check. The logs showed no obvious error. After digging in, the root cause was dependency isolation — dotenv and several other packages were installed at the project root but not inside the backend/subfolder. Since the backend runs from its own directory, it couldn't find its dependencies and crashed on import. Fixed by adding an explicit cd backend && npm install step to the build pipeline so the backend has its own complete node_modules.

AI story generation consistency and quality control

Getting the AI to generate stories that were genuinely funny, age-appropriate, and consistent across 10 pages required significant prompt engineering. Early outputs were either too generic ("Biscuit went on an adventure and had fun") or tonally inconsistent (page 3 would be whimsical, page 7 would read like a Wikipedia article). We solved this by structuring the generation into a two-pass system: first pass generates a story outline with character arc and tone guidelines, second pass writes each page while referencing the outline. This kept the narrative coherent and the humor consistent throughout.

Quiz-to-story pipeline latency

The full generation pipeline (quiz intake → story generation → illustration generation → PDF assembly) initially took 45+ seconds, which felt like an eternity for a user staring at a loading screen. We broke this into a streaming experience: the story text generates first and displays page-by-page while illustrations render in the background. Users start reading immediately while the full book assembles. Perceived wait time dropped from 45 seconds to under 10.

Print-ready PDF formatting

Digital PDFs and print-ready PDFs have completely different requirements — bleed margins, CMYK color space, spine width calculations for hardcover, DPI requirements for illustrations. A PDF that looks perfect on screen can print with clipped edges and washed-out colors. We built a separate export pipeline that re-renders the entire book at 300 DPI with proper bleed margins and print-safe color profiles, triggered only when a customer selects the softcover or hardcover tier.