Mochi Reader

Why I built it

Most of what I want to read online is in a language I don't speak well: Korean and Japanese comics, Chinese web novels, news and forums. Copying text into a translator breaks the flow, and text inside images can't be copied at all. Mochi Reader is a browser that translates the page while you read it.

Features

• Translated inside the image. Mochi finds the text in each picture, groups it into speech bubbles and redraws the translation to fit, keeping the bubble's colors. Hold the page to peek at the original.
• Works on any website. It's a full browser: paste a link or search, log in when a site needs it. Pages are translated two screens ahead, so the text is ready when you scroll there.
• Web novels and articles too. Turn on page translation and every paragraph changes as you scroll.
• PDFs and photos. Open a PDF from Files or iCloud Drive, or pick pages from your photo library.
• Private by design. Text recognition and Apple Translation run on the device. Images are never uploaded, and no account is needed.
• Translates Korean, Japanese, Chinese and English into English, Vietnamese, Japanese, Korean, Chinese, Thai, Indonesian and French.

How I built it

• iOS and macOS: SwiftUI, one codebase for iPhone and Mac. WKWebView with an injected script that finds images and text on the page.
• On-device OCR: Apple Vision plus a text-detection model running on ONNX Runtime, then Apple Translation.
• Android: Kotlin + Jetpack Compose, sharing the same injected script.
• Backend: Firebase Auth (Sign in with Apple / Google), Firestore, Cloud Functions; StoreKit 2 subscriptions with a 3-day free trial and invite rewards.
• Web: Astro landing page in 12 languages, admin console in Vite + React, deployed with Docker.

Free on the App Store for iPhone, with Pro plans for unlimited pages. The Mac version is on its way.