LiveCream & Slush
A retro soda fountain for your Ninja Slushi and Creami.
Tell it what is in your fridge and it slings you a frozen drink. Retro soda-fountain vibes, 160+ recipes, real freeze-and-booze science baked in.

Why I built it
I got a Ninja Slushi and immediately hit the classic problem: I had a machine, a fridge full of random stuff, and no idea what I could actually make. Every recipe online was scattered, none scaled to my machine size, and nobody warned you when a mix would not freeze or was way too boozy to slush. So I built the site I wished existed, a fun retro soda-fountain spot where you tell it what you have got on hand and it tells you what you can make right now.
What it does for people
- A What Can I Make ingredient finder. Pick what is in your fridge and it instantly shows buildable recipes, plus an almost-there list when you are one or two short
- 161 frozen-drink recipes, from sodas and mocktails to milkshakes, frappes, sorbets, gelato and frozen cocktails, searchable by name or ingredient
- Automatic machine scaling for all four Slushi sizes (48, 64, 74, 112 oz) so every quantity is right for your machine
- Live safety warnings baked into every recipe. It flags freeze-risk mixes and over-the-limit alcohol before you pour
- A 21+ age gate on boozy recipes, plus community submissions and cherry-rating reviews so the site keeps growing itself
Under the hood
- Turned a corrupted research PDF into a clean product. I extracted a mangled JSON block full of span artifacts and per-token line breaks, bracket-matched it back to valid data, then normalized 86 ingredients into pickable categories
- Set-theory match logic. A recipe counts as buildable only if you have all its core ingredients, optional ones ignored, which is what makes the finder feel smart
- A real food-science engine: machine scaling multipliers, a spirit-to-ABV lookup, a hard-spirit volume limit, and freeze-risk detection
- Static-first architecture. Recipes ship as JSON static props with client-side search via useMemo, so there is no database and no API to slow things down or cost money
- SEO done properly: Recipe JSON-LD for rich results, sitemap, OpenGraph, plus auto-generated recipe images and Pinterest pins from scripts
The cool bits
- The whole thing is dressed up as a 1950s soda fountain: Pacifico script headings, candy stripes, polka dots, scalloped corners and chunky diner drop shadows. It has got personality
- You can walk up with literally whatever is in your fridge and it hands you a drink you can make in the next five minutes
- It quietly does real chemistry. Most people have no clue their mix will not freeze until it fails, and this catches it up front
- It even auto-generates its own share images and Pinterest pins. The marketing engine is part of the build
How long it took
About two weeks of focused building, from a corrupted research PDF to a full monetization-ready site.
2 weeks · 1 sprint · 112 commits
What I worked through
The trickiest part was the data recovery. The source recipe database came as a mangled PDF with corrupted JSON, so I had to extract, strip, and bracket-match it back into something usable, then normalize dozens of ingredient synonyms so the finder would not treat Cold Water and Water as different things. The other big one was the alcohol angle. 21 of the original 30 recipes were boozy, which complicates ad-network eligibility, so I designed around age-gating, safety flags, and a monetization plan that starts on non-alcohol pages.