AI Translator

Select text on a webpage, right-click, and watch the translation appear in a floating panel—without switching tabs or copying and pasting. AI Selection Translator is an open-source Chrome extension built on Manifest V3. Translate selected text, page regions, and text inside images into your preferred language using your own OpenAI, DeepSeek, or DeepL API key. ▍Four ways to translate • Selected text Select text and choose “Translate to…” from the right-click menu. The translation appears in a centered panel, with streaming output where supported and an expandable original-text view. Selections in any language—including Chinese—can be translated. Selections containing only numbers, punctuation, or emoji do not trigger the translation offer. Change the target language directly in the panel to translate the original selection again. Save it as your default only when you choose to. • Page regions Add a translation button to a specific area of a webpage. Click it to replace the text in that region with its translation, and restore the original when needed. Rules are saved by domain, so the button returns when you revisit the website. Longer regions are processed in batches, with translations appearing progressively. • Images Right-click an image to recognize its text and overlay translations at the detected positions. Long images are split into smaller sections for recognition. Changing the target language reuses the recognized text, so the same image does not need to be scanned again. Image recognition requires a vision-capable provider and model. • Translation panel Open a dedicated page with source text on the left and translation on the right. Use it for longer passages or text from outside the browser. Choose from 24 target languages: Simplified Chinese, Traditional Chinese, English, Spanish, Portuguese, French, German, Italian, Dutch, Russian, Ukrainian, Polish, Turkish, Arabic, Hebrew, Persian, Hindi, Bengali, Urdu, Indonesian, Vietnamese, Thai, Japanese, and Korean. Detect the source language automatically or select it explicitly. Source language, target language, and interface language are independent settings. Provider support varies by language and feature. ▍Designed for everyday reading • Stay on the page: translate while reading, without interrupting your workflow. • Streaming output: translations appear progressively where supported. • Light and dark themes follow your system preference. • Drag the translation panel to reposition it, or press Esc to close it. • Keyboard shortcut: ⌘⇧Y on macOS or Alt+Shift+T on other platforms. • Localized interfaces for 24 languages, including right-to-left layouts. • Choose your default translation language independently of the interface language. Experimental video subtitle translation is also available as an optional feature. Speech recognition support depends on your browser, device, and source language. ▍Privacy • No developer-operated translation server. Translation requests go directly to the provider or custom endpoint you configure. • No analytics, tracking, or advertising. • Your API key is stored locally in chrome.storage.local and is not synced through your Chrome account. • Content is sent for translation only when you actively request it. The developer does not receive your translation history. • Video speech recognition can run locally on supported devices. Optional cloud recognition sends audio to the browser’s speech recognition service. • The complete source code is available under GPL-3.0 for inspection. ▍Getting started 1. After installation, the settings page opens automatically. Choose a provider, enter your own API key, and use “Test connection” to check your configuration. 2. Choose your interface language and default translation target. 3. Select text on a webpage, right-click, and choose “Translate to…” to begin. Requires Chrome 116 or later. Experimental video features may require a newer browser and additional setup. This version uses your own provider credentials. Any API charges are billed by your chosen provider. ▍Known limitations • Browser-protected pages, including chrome:// pages and the Chrome Web Store, do not allow content-script injection. Translation overlays are unavailable there. • Individual text translations are limited to 6,000 characters. Longer selections are truncated, with a notice shown in the panel. • Custom API endpoints require separate permission for their domain and must use HTTPS. • Available languages, image recognition, and streaming capabilities depend on your provider and model. ▍About this project An independent, open-source project. It is not affiliated with, endorsed by, or sponsored by OpenAI, DeepSeek, DeepL, or their associated companies. ChatGPT, DeepSeek, DeepL, and other product names are trademarks of their respective owners and are used solely to identify compatible services. Website: https://huaci.app/ Source code and issue reports: https://github.com/csswork/AiTranslation Privacy policy: https://github.com/csswork/AiTranslation/blob/main/PRIVACY.md