ImageTool
Creator & Full-Stack Architect · 2026

ImageTool

A high-performance Neo-Brutalist background removal SaaS built with Next.js that runs machine learning inference entirely in the user's browser using WebAssembly. Eliminates expensive GPU server bills and reduces manual graphic editing workload by 70% while bypassing serverless payload limits via direct cryptographically signed uploads.

Client
Creative Dev / SaaS Tooling
Stack
Next.js 16, TypeScript, WASM, Tailwind CSS v4, Cloudinary, MongoDB, JWT
Year
2026
Role
Creator & Full-Stack Architect
ImageTool Preview
Challenge

Traditional background removal tools rely on server-side GPU instances, charging users steep per-image fees or incurring unsustainable server bills. Furthermore, serverless platforms like Vercel enforce strict 4.5MB request payload limits on API routes, causing uploads of modern high-resolution photos (8MB–12MB) to fail with HTTP 413 Payload Too Large errors.

Architecture

Engineered a client-side execution architecture using @imgly/background-removal powered by the isnet_quint8 WASM model running on client Web Workers with hardware-accelerated WebGL shaders. Solved the serverless payload barrier by implementing a direct-to-Cloudinary signed upload pipeline: the client authenticates via JWT, requests a short-lived cryptographic signature from a lightweight JSON endpoint, and streams the processed image directly to Cloudinary storage.

Approach

Engineered client-side WebAssembly background removal running offline in browser Web Workers. Implemented cryptographic signature generation endpoints allowing users to stream up to 12MB photos directly to Cloudinary, bypassing Vercel's 4.5MB limit. Designed a distinct high-contrast Neo-Brutalist aesthetic (#ffe17c yellow, #b7c6c2 sage, #171e19 charcoal) with hard mechanical button interactions.

Outcome

Reduces manual editing workload by 70% for creative professionals and designers. Delivers $0 server compute costs by executing 100% of ML inference locally, while supporting direct signed uploads up to 12MB with zero serverless payload drops.

Core Capabilities
70% Workload Reduction

Automates background removal in seconds with one-click asset export.

100% In-Browser WASM

Executes ML models locally in Web Workers with zero server GPU costs.

Signed Direct Uploads

Cryptographic signatures allow direct 12MB uploads, bypassing Vercel 4.5MB limits.

Neo-Brutalist UI

High-contrast aesthetic with hard mechanical shadows and physical push buttons.

ImageTool Screenshot 1
ImageTool Screenshot 2
Results & Verified Metrics
70%
Workload Reduction
$0
Server Compute Cost
12MB
Max Direct Upload Payload