HotLingo AI Translation
A multi-platform AI translation product that turns instant Vietnamese ↔ English translation into a 2-3 second action from any app — no context switching, no API key.
/ Overview
HotLingo is a multi-platform AI translation product designed for Vietnamese users who work with English every day. A native macOS menu-bar app — with Windows, a web/PWA companion, and an admin console — lets users translate selected text or a screenshot region with a single hotkey, streaming the result token-by-token right next to the cursor. A 3-layer context system keeps translations natural, and a pay-as-you-go credit model removes the two biggest local barriers: API keys and international payment cards.
Get in touch/ Project Details
Name:HotLingo AI Translation
Client:HotLingo
Industries:AI / Translation, Productivity
Date:2026
Project page:https://hotlingo.vn
/ Challenge
The goal was to make AI translation effortless for a market blocked by technical friction — no API keys, no foreign cards, no app-switching. That meant a truly system-wide translator that reads text or screen regions from any app, streams results instantly, stays accurate through context, and bills in a mental model Vietnamese users already trust — topping up credit, not a monthly subscription.
/ Provided Services
Translate Anywhere (⌃⌥T)
Select text in any app and a floating panel streams the translation next to the cursor — powered by global Carbon hotkeys and the macOS Accessibility API, with a clipboard fallback.
Screenshot OCR → Translate (⌃⌥S)
Drag to select any screen region; Apple Vision OCR extracts the text (Vietnamese included) and translates it instantly — no screenshot files left on disk.
3-Layer Smart Context
Persistent, per-message, and screenshot context layers merge into each request so translations read naturally, with tone and domain preserved.
Pay-as-you-go Credits
Three modes coexist — free Google Translate, credit-based AI, and bring-your-own-key — with VietQR top-up so users pay without an international card or an API key.
/ Solution
A native-first architecture keeps the app tiny and instant, while a model-agnostic backend keeps AI flexible and costs under control. SwiftUI + AppKit deliver a ~10MB zero-dependency macOS binary; Supabase handles auth, credits, and Edge Functions that proxy to OpenRouter so the AI model is a single admin-changed string; SSE streams every translation token-by-token across native, web, and mobile surfaces.
/ Tech stack
SwiftUI + AppKit
Native macOS menu-bar app — ~10MB binary, zero external dependencies, notarized.
Apple Vision + Carbon + Accessibility API
Offline OCR, global hotkeys, and reading selected text across every app.
Supabase
Authentication, credit-based billing, and Edge Functions for the AI proxy.
OpenRouter
Model-agnostic AI routing — swap GPT-4o-mini, Gemini or Claude via one config string.
Next.js + next-intl
Localized marketing landing page and admin console.
React + Vite + Capacitor
Web/PWA and mobile companion app with SSE real-time streaming.
