Immersive Translation#

Didn’t expect the second blog post on the new site to still be self-promotion—new tech in the next one 😈

Preface#

Late last year, when Deepseek v3 was released, I made an immersive translation plugin as a little celebration. Unexpectedly, over half a year later, “Immersive Translation” drama popped up. So I dusted off the old plugin, tweaked it, and shipped it—you can check it out here ↗.

The main goals of this plugin are to be fully open source and to use only your own API. I’ll add more features when I have time, such as custom hotkeys.

Changelog#

V 2.1#

Supports Chrome AI; entirely on-device.

V 2.0#

Added new features

• Multilingual UI (i18n) – English, Chinese, Japanese, Korean, French, German, Spanish.
• Reasoning view (🧠) – If the model returns <think>…</think>, a button appears to expand and view it on demand.
• YouTube subtitles – Bilingual overlay; choose Built-in or API first.
• PDF & images – Use local Tesseract.js OCR to extract text, with fallback to LLM vision when needed.

V 1.0#

• Inline page translation – Inject the translation directly beneath each paragraph/list/heading (no giant popups).
• Full-page translation – Supports full-page translation, automatically detects the main content, and reduces API requests.
• Dictionary bubble – When the selected text is a single word, show a concise learning entry.