UGTBrowser

Inline translation of highlighted text using state-of-the-art AI models (OpenAI, Anthropic Claude, Google Gemini).

Universal Game Translator (Browser extension edition) Why did I make this? Because the default translate option in Chrome makes MANY MISTAKES with Japanese to English. Using the best AI models lets you choose quality over speed - when researching old game blogs and such, it's totally worth it. I use a text-fragment approach that lets the LLM understand the full context (all fragments are sent at once and in order), but still allows us to do a streaming translation with in-place text replacement that doesn't mess up links and images. WARNING: This extension is free and open source, but the LLMs actually cost money and generally make you put in a credit card to get an API key and charge per use. So just be aware that it can add up. Try Gemini, for example; the first million tokens per month are free. Features: - Translate selected text using the latest LLM models (OpenAI, Anthropic, Google) - In-place image translation - intelligently convert an image to use a different language with a flip control to compare original and translated - Video-frame translation - translate a paused frame on YouTube and other video pages - Text-to-Speech - Have text read aloud via ElevenLabs or Google Cloud TTS - Create Lesson - Generate detailed language breakdowns with readings, meanings, and grammar notes - Follow-up Chat - Ask questions about the translated content - Support for a secondary "Creative task", by default is explains cultural notes - Securely stores your API keys and settings locally in your browser - Nothing is sent out except the text to the LLM providers you're using - Customizable prompt templates for advanced users - Context menu integration for easy access - During translation, a little animation is shown with an option to see the text coming in - There is a way to see the last prompt/response pair in the settings, good for debugging New in V2.3.0: New in V2.3.0: * Updated to the newest OpenAI, Anthropic and Google models (older ones are still selectable) * Image and video-frame translation use OpenAI's newer image model, with a slower, higher quality option in the new Image Model setting * Added the newest ElevenLabs text-to-speech models * Fixed translation doing nothing on pages that put their text inside a shadow DOM, like Yahoo! Auctions item descriptions * Fixed notes and panels getting cut off in collapsed YouTube comments, and only part of a multi-line YouTube comment getting translated * Misc. fixes SAFETY: I am Seth A. Robinson, a game developer with over thirty released games and creator of many freeware apps including Universal Game Translator and Universal Game Translator: Live. This app does zero tracking or phoning home, it only communicates with the LLMs you setup. (required for translation) The full source code (MIT license) is on github, linked below.