Local Browser Translator Autor: underlines
Translates web pages in place using your own local or cloud LLM — llama.cpp, Ollama, vLLM, LM Studio or any OpenAI-compatible endpoint. Built so a slow local model still feels fast.
Metadáta rozšírenia
Snímky obrazovky
O tomto rozšírení
Local Browser Translator translates web pages through an LLM endpoint you control — not a translation API run by us. Point it at a local model (llama.cpp, Ollama, vLLM, SGLang, LM Studio, llama-swap) or a cloud OpenAI-compatible provider (OpenAI, OpenRouter, Groq, DeepSeek, Together, and more), and it translates the page in place, in your target language.
It's built around one idea: local LLMs are slow, and a single GPU can't be parallelised across tabs. So instead of firing hundreds of tiny requests and leaving you staring at a half-translated page, it:
Your data stays where you point it. The extension talks only to the endpoint you configure. The default is a loopback local server (http://localhost:8080/v1); if you configure a remote endpoint, page text goes there and nowhere else. There is no telemetry, no analytics, and no code loaded from anywhere but the extension package itself.
Site access is opt-in, per site. The extension does nothing on install. You enable it for a site with one click, and even then translation only starts when you ask for it (button, hotkey, or the in-page control).
Works with: llama.cpp, Ollama, vLLM, SGLang, LM Studio, llama-swap, and OpenAI-compatible cloud APIs (OpenAI, OpenRouter, Groq, DeepSeek, Together, and others that speak the same protocol).
Keyboard shortcuts: translate page, revert page, translate selection, and toggle how translations are displayed (replace / bilingual / hover-reveal) — all rebindable in about:addons.
It's built around one idea: local LLMs are slow, and a single GPU can't be parallelised across tabs. So instead of firing hundreds of tiny requests and leaving you staring at a half-translated page, it:
- Translates what you're looking at first. Above-the-fold and main content are prioritised; footers, ads and legal boilerplate go last.
- Streams results as they're produced. Translated segments appear as soon as the model finishes them — you don't wait for the whole page.
- Never translates the same text twice. A local translation memory means repeated headers, menus and boilerplate across pages on the same site cost nothing after the first hit.
- Degrades instead of failing. Malformed model output gets repaired, retried smaller, or the original text is left in place — the page is never left half-mangled.
Your data stays where you point it. The extension talks only to the endpoint you configure. The default is a loopback local server (http://localhost:8080/v1); if you configure a remote endpoint, page text goes there and nowhere else. There is no telemetry, no analytics, and no code loaded from anywhere but the extension package itself.
Site access is opt-in, per site. The extension does nothing on install. You enable it for a site with one click, and even then translation only starts when you ask for it (button, hotkey, or the in-page control).
Works with: llama.cpp, Ollama, vLLM, SGLang, LM Studio, llama-swap, and OpenAI-compatible cloud APIs (OpenAI, OpenRouter, Groq, DeepSeek, Together, and others that speak the same protocol).
Keyboard shortcuts: translate page, revert page, translate selection, and toggle how translations are displayed (replace / bilingual / hover-reveal) — all rebindable in about:addons.
Hodnotené 0 od 0 recenzentov
Povolenia a údaje
Požadované oprávnenia:
- Pristupovať ku kartám prehliadača
Požadovaný zber údajov podľa vývojára:
- Obsah webovej stránky
Ďalšie informácie
- Odkazy doplnku
- Verzia
- 1.2.0
- Veľkosť
- 308,26 kB
- Posledná aktualizácia
- pred 5 dňami (1. okt 2026)
- Príbuzné kategórie
- Licencia
- Licencia MIT
- Zásady ochrany osobných údajov
- Prečítajte si Zásady ochrany osobných údajov pre tento doplnok
- História verzií
- Pridať do kolekcie