SuryanshChourasia
All work

Mine. Live, with paying users · 2025

Imagika

I built and launched an AI product photography tool that turns one reference image into category-aware campaign visuals in under a minute.

Frontend developmentInterface design in codeFull-stack AI product

Small e-commerce teams often need more campaign imagery than their photography budget can produce.

I built Imagika to test whether one reference upload could become a useful set of commercial visuals. The difficult part was not generation alone. The product also needed category-specific treatment, clear choices, dependable credits, and a workflow people would pay to repeat.

Imagika running on a Studio Display in a warm sunlit studio, generated product photography beside the source reference image.
Imagika's how-it-works section: 'Select your favorite style, drag and drop your photo', an upload box with a Generate button, and two example looks either side joined by a dashed path.
The whole product in one screen: pick a look you love, drop in your photo, press Generate.
'Try It On Instantly Anywhere You Shop': an Extension pill on a dashed path running through Amazon, H&M, Zara and eBay logos, with a Get Extension button.
The browser extension, drawn as one path through the stores where people actually shop.
'Smarter Tools Seamless Creation' above three cards: Social Trends photos, a Design and Styles node linked to social icons, and AI Models with AI Product jewellery shots.
Three tools, three cards: social trends, design styles for every platform, and AI models for product shots.
'Trusted Payment Partners' with a large Dodo Payments badge and UPI, Visa and MasterCard logos, each with a soft coloured label.
Payments people in India recognise, presented with the same playful pills as the rest of the site.
A soft blue card reading 'Create Post and Tag us on your Socials' with a Get Started button and floating #imagikacreation tags.
The page closes by asking people to share what they made.

The decisions that made it work

01

Detect the category first

The prompt layer is category aware. Footwear, apparel, jewellery and packaged goods each need a different realism treatment, so the system classifies before it generates rather than sending everything through one prompt.

02

Four variations, not one

People do not want the right answer, they want to choose. Four styles per generation turned out to matter more than any single quality improvement.

03

Four front doors, one backend

Web app, design canvas, brand photoshoot wizard and a browser extension, all against a single Hono API mounted inside Next.js.

04

An append only credit ledger from day one

Billing that mutates a balance column will eventually lie to you and you will not know when it started. Every debit and credit is a row. The balance is derived. It cost an extra day and has never once been wrong.

1,500+

users

30 to 60s

reference to result

7

auto-detected categories

4

surfaces, one backend

Built withNext.jsHonoDrizzlePostgresGeminiCloudinaryDodo PaymentsBun
Next projectMathmatikaMine
Want one like this?Landing pages from $999, live in about 10 days.Start a project