Bedrock Translator

Bedrock Translator translates the web page you are viewing - or just the text you select - using your own Amazon Bedrock account. Your browser talks directly to Amazon Bedrock: no intermediary server, no data collection. FULL-PAGE TRANSLATION - One click translates the entire page in place, keeping the original layout. - Progressive rendering: translated sections appear immediately as they complete, with a small progress pill in the corner instead of a blocking overlay. - Restore the original page with one click. SELECTION TRANSLATION - Select text, right-click, and choose "Translate Selection". - The translation streams token-by-token into an inline tooltip right below your selection. AUTOMATIC MODEL DISCOVERY - The models available to your Bedrock account are discovered automatically; new model generations appear without an extension update. - Auto mode (default) picks the newest efficient model per provider (currently GPT-6 Luna and Claude Sonnet 5) and keeps reasoning off, which is what translation needs. - Manual mode lets you pick any discovered model, including Grok 4.6. - Select 2 or more and they split the page into chunks and translate them in parallel using work-stealing scheduling (faster models automatically take more chunks). 5 TARGET LANGUAGES - Korean (default), English, Japanese, Simplified Chinese, Spanish. DESIGNED TO STAY OUT OF YOUR WAY - Minimal monochrome UI with automatic light/dark theme. - Adjustable parallel speed (1-5) for API rate-limit control. - Translation cache: the last 5 pages are cached locally for instant re-translation without API calls. REQUIREMENTS - An Amazon Bedrock API Key (us-west-2 region) with access to the models you use. Model discovery also needs bedrock:ListFoundationModels and bedrock:ListInferenceProfiles; without them the built-in default models are used. PRIVACY - Your API key and preferences are stored only in your browser (Chrome sync storage). - Page text is sent only to Amazon Bedrock for translation and is not stored by this extension. Model discovery calls contain no page content. - No analytics, no tracking, no third-party servers.