DEV Community: Suman Gorai

I Built an AI-Powered Virtual Try-On Tool for Shopify — Here’s How It Works (Torziva)

Suman Gorai — Tue, 24 Mar 2026 12:41:03 +0000

Online shopping is broken in one major way — you can’t try before you buy.

Customers constantly ask:

“Will this look good on me?”
“Will it fit my body?”
“Is this worth buying?”

This uncertainty leads to low conversions and high return rates.

So I decided to build something to solve this.

👉 Meet Torziva — an AI-powered virtual try-on tool for Shopify.

🚀 The Idea Behind Torziva

The goal was simple:

Let users upload their photo and instantly try clothes virtually.

Instead of imagining how a product might look, users can see it on themselves.

🧠 How It Works (User Flow)

Here’s the basic flow:

User visits a Shopify product page
Clicks on “Try On”
Uploads their photo
AI processes the image
Output: realistic try-on preview

This entire process takes just a few seconds.

⚙️ Tech Stack & Architecture

Building Torziva wasn’t easy. Here's a simplified version of the stack:

Frontend: Embedded Shopify UI (App Extension)
Backend: Node.js / API-based system
AI Processing: Image-to-image transformation models
Storage: Cloud-based image handling
Integration: Shopify App + API

🧩 The Core Challenge

The hardest part wasn’t coding.

It was:

👉 Making the output look realistic

Problems I faced:

Clothes not aligning with body
Lighting mismatch
Distorted outputs
Slow processing time

Solving these required:

Prompt tuning
Image preprocessing
Model selection & testing

⚡ Performance Optimization

Speed matters.

Users won’t wait 30 seconds just to try a t-shirt.

So I optimized:

Image size before processing
API response time
Queue handling

Goal: < 5 seconds result

💡 Why This Matters for Shopify Stores

Torziva helps store owners:

📈 Increase conversions
🔁 Reduce returns
😄 Improve user experience
🛍️ Boost engagement

This is especially powerful for:

Fashion brands
D2C stores
Custom clothing businesses

🔮 What’s Next

I’m currently working on:

Multi-angle try-on
Video-based try-on
Better realism with advanced models
Mobile optimization

🙌 Final Thoughts

Building Torziva has been an incredible journey.

From idea → prototype → working SaaS, I learned:

AI is powerful, but not plug-and-play
UX matters more than tech
Speed is everything

If you’re building in AI + e-commerce, this space is just getting started.

🔗 Try Torziva

If you’re curious, you can check it out here:

👉 https://www.torziva.site/

I’d love your feedback 🙌

🏷️ Tags

#ai #shopify #saas #webdev #startup #machinelearning

I Tried to Build a Startup Without a Team --- Here's What Happened

Suman Gorai — Sat, 21 Mar 2026 04:25:59 +0000

🚀 Intro

Everyone talks about building startups with a team.\
But I wanted to try something different.

What if I build everything alone --- product, backend, frontend,
marketing... everything?

This is the story of what actually happened.

💡 The Idea

I wanted to build a simple SaaS product that helps freelancers create
invoices and get paid easily.

No big funding.\
No team.\
Just me, my laptop, and a lot of ideas.

⚙️ The Stack I Chose

To keep costs low, I used:

Next.js (Frontend)
Supabase (Backend + Auth + DB)
Vercel (Deployment)
PayPal (Payments)

Everything was either free or very low cost.

😵 The Reality Hit Me

Building alone sounds cool... until you actually do it.

I had to fix bugs and design UI
Handle backend logic and database issues
Think about marketing and user acquisition
Debug things at 2 AM with no one to help

There were days I felt completely stuck.

📉 Biggest Mistakes I Made

I started building without validating the idea\
I ignored marketing until the product was "perfect"\
I tried to do everything instead of focusing on what matters

These mistakes slowed me down a lot.

🔥 What Actually Worked

Keeping the product simple
Using free tools to reduce risk
Learning by building instead of overthinking
Sharing progress publicly

That's what kept me going.

🎯 Final Thoughts

Building a startup alone is possible... but it's not easy.

You'll feel stuck.\
You'll doubt yourself.\
But you'll also learn faster than ever.

If you're thinking of building something solo --- just start.

You'll figure things out along the way.

📢 CTA

If you're building something solo, I'd love to hear your story.

Drop it in the comments 👇

🚀 I Built an AI Virtual Try-On for eCommerce (To Fix a Huge Conversion Problem)

Suman Gorai — Thu, 19 Mar 2026 03:34:12 +0000

Hey devs 👋

I’ve been working on something interesting lately — an AI-powered virtual try-on tool for eCommerce stores.

🤔 The Problem

One thing I kept noticing while exploring online stores:

Customers hesitate to buy because they can’t visualize how a product will look on them.

This leads to:

❌ Low conversion rates
❌ High return rates
❌ Poor user confidence

And honestly… it makes sense.

💡 The Idea

So I thought:

What if users could instantly try products using AI, directly from the product page?

No apps. No friction.

Just click → try → decide.

🛠️ What I Built

An AI virtual try-on system that:

Lets users try products in real-time
Works directly inside eCommerce stores
Requires minimal integration
Focuses on speed and simplicity

⚙️ How It Works (Simple Overview)

User visits a product page
Clicks "Try Now"
AI processes the input
Shows a realistic preview

All within seconds.

🎯 Who It’s For

Shopify store owners
D2C brands
Fashion & accessory sellers
Anyone struggling with conversions

📈 Goal

The main goal is simple:

Help store owners increase conversions and reduce returns using AI.

🔥 What I Need From You

I’m still improving this, so I’d love feedback from the dev community 🙌

What features would you add?
Any ideas to make it more useful?
What would you improve technically?

🌐 Try It Out

If you’re curious, you can try it here:
👉 https://torziva.site

Thanks for reading!
Would love to hear your thoughts 🚀

How I Fixed a Silent 500 Error in My Supabase Edge Function

Suman Gorai — Wed, 18 Mar 2026 03:15:24 +0000

How I Fixed a 500 Error That Was Silently Killing My SaaS

I launched Torziva — an AI virtual try-on
tool for fashion stores — and users were getting 500 errors on
every try-on attempt.

The weird part? Supabase logs showed nothing. Completely blank.

The Symptom

Browser console: Failed to load resource: 500
Supabase edge function logs: blank
fal.ai dashboard: requests weren't even arriving

Classic "where is the request dying?" mystery.

The Investigation

Step 1 — Checked Vercel env vars

My edge function was hosted on Supabase but the request
originated from Vercel. Turned out FAL_API_KEY was missing
from Supabase secrets — it was only in Vercel.

Added it to Supabase → still failing.

Step 2 — Checked the actual error

Added more granular try/catch logging and finally saw it:

Error: Processing timeout. Please try again.

Step 3 — Found the real culprit

My polling loop was set to 60 attempts × 2000ms = 120 seconds.

Supabase edge functions have a default timeout of 60 seconds.

The function was dying before fal.ai even finished processing.

The Fix

Two changes in the edge function:

// Before
const maxAttempts = 60;
num_inference_steps: 30,

// After  
const maxAttempts = 25;  // 25 × 1500ms = ~37 seconds
num_inference_steps: 20, // faster generation

Also changed polling interval from 2000ms to 1500ms.

Total processing time dropped from ~90s to ~35s.
Now well within Supabase's limit.

The Lesson

If your Supabase edge function logs are blank, the function
isn't crashing — it's timing out before it can log anything.

Always check:

Are your secrets in Supabase, not just Vercel?
Is your total execution time under 60s?
Are external API calls (fal.ai, OpenAI, etc.) taking longer than expected?

Stack

Supabase Edge Functions (Deno)
fal.ai IDM-VTON for try-on generation
Next.js 14 + Vercel frontend

If you're building anything with fal.ai or Supabase edge
functions, happy to answer questions in the comments.

And if you run a fashion store — this is what powers
torziva.site 👕

How I built an AI virtual try-on button for fashion stores using fal.ai and Supabase

Suman Gorai — Tue, 17 Mar 2026 03:53:06 +0000

The problem I was trying to solve

Fashion ecommerce has a dirty secret — 30-40% of all purchases get
returned. The #1 reason? "It didn't look right on me."

Shoppers can't tell how a dress will fit their body from a flat
product photo. So they buy, try, and return.

I built Torziva to fix this.

What Torziva does

Torziva adds a "Try On" button to any fashion store's product page.
When a shopper clicks it:

They upload their photo
AI places the exact garment on their body
They see a realistic on-model photo — in under 30 seconds

Real fabric texture. Real fit. Real colors. On their actual physique.

The tech stack

Here's what I used to build it:

fal.ai — for the AI image generation / virtual try-on model
Supabase — auth, database, storage
Next.js 14 — frontend + API routes
Vercel — deployment

The core flow is simple:

User uploads photo → 
API sends to fal.ai with garment image → 
fal.ai returns generated image → 
Display result to user

Why fal.ai?

I evaluated a few image generation APIs. fal.ai won because:

Fast inference (~30 seconds for try-on)
Pay per generation ($0.075/image)
Good try-on model quality for fashion garments
Simple API, easy to integrate

The Supabase setup

I used Supabase for three things:

Auth — email + Google OAuth for store owners
Database — storing generation history, store configs
Storage — saving generated images

Supabase MCP made the setup surprisingly fast.
Connected directly from Cursor and scaffolded the schema in minutes.

Challenges I ran into

1. Garment detection quality
The AI works best with flat lay photos on white backgrounds.
Dark backgrounds or heavy shadows confuse the model.
Had to add input validation and guidance for store owners.

2. Shopify integration
Getting the Try On button to appear correctly on Shopify product
pages required careful theme injection. Shopify's script tag API
made this manageable but took longer than expected.

3. Pricing
At $0.075/generation, pricing the product correctly was tricky.
Had to account for Shopify's 15% fee on top.
Ended up with a usage-based model that scales with store size.

Results so far

Launched on Product Hunt on March 12
Listed on Peerlist, BetaList
First users onboarding now

Still early days but the feedback from fashion store owners has
been positive — especially the "no photoshoot needed" angle.

What's next

Full Shopify app (in review)
Support for more garment types
Bulk generation for store catalogs

If you're building anything in fashion tech or ecommerce AI,
would love to connect. And if you run a fashion store —
try it free at torziva.site.

Happy to answer any questions about the tech stack in the comments 👇