BabelFishAI by jls42.org 作成者: Julien LS

Babel Fish AI - AI-Powered Voice Transcription and Translation Extension

To use the extension, you will need an API key from one of the supported providers:

Provider Get an API key
Mistral AI Mistral AI: console.mistral.ai/api-keys
OpenAI OpenAI: platform.openai.com/account/api-keys
Gemini Gemini (Google): aistudio.google.com/apikey
🚅 Custom/LiteLLM: To use your own API endpoints
Babel Fish AI is an innovative browser extension designed to provide powerful voice transcription with multi-provider support. Turn your voice into text with remarkable accuracy using the transcription APIs from Mistral AI (Voxtral), OpenAI (gpt-transcribe, Whisper), or Gemini (Gemini 3.5 Transcribe), with optional real-time automatic translation. You can use Babel Fish AI exclusively for transcription or enable on-the-fly translation as needed.

🌟 Features
Advanced Voice Transcription

High-quality audio capture through your device's microphone.
Accurate transcription via the Voxtral (Mistral AI), gpt-transcribe and Whisper (OpenAI), or Gemini 3.5 Transcribe (Google) APIs.
Multi-provider support: freely choose between Mistral AI, OpenAI, Gemini, or a custom endpoint.
Multilingual support for speech recognition and text display, allowing you to transcribe voice input in different languages and display the results (transcription and translation, if enabled) in the language of your choice.
Automatic text insertion into the active field or display in a dedicated dialog box.
Intelligent Translation and Rephrasing

Immediate translation of transcriptions into various languages, enabled as needed.
Text rephrasing to improve style and clarity.
Use of an advanced AI model to ensure translation faithful to the original meaning.
Freedom to use transcription exclusively or combine transcription and translation.
Powerful Context Menu

"Rephrase selection" option to instantly improve your selected text.
"Translate selection" option with a submenu containing all available languages.
"Correct spelling" option to correct spelling, grammar, and punctuation errors.
Direct replacement of selected text with its translated, rephrased, or corrected version.
Seamless integration into the browser's native user interface.
Intuitive and Customizable User Interface

Flexible display mode: active input field or floating dialog window.
Configurable status banner with options for color, opacity, and display duration.
Keyboard shortcut (Ctrl+Shift+1 or ⌘+Shift+1 on Mac) to start/stop recording.
In Firefox, a “priority shortcut” for sites whose page editor intercepts the key combination (ChatGPT, Notion, etc.): enabled upon installation and can be disabled in the options (see PRIVACY.md).
"Keep open" option to control how long results are displayed.
Custom icon incorporating a microphone and the number "42" for instant recognition.
Advanced Options

Multi-provider support: Mistral AI, OpenAI, Gemini, and Custom/LiteLLM for maximum flexibility.
Ability to customize transcription and translation models for each provider.
Available OpenAI models: GPT-4o mini (default), GPT-4.1 (mini/standard), GPT-5.4 (nano/mini/standard), and GPT-5.6 (luna/terra/sol). Transcription: gpt-transcribe (default), whisper-1, gpt-4o-mini-transcribe, and gpt-4o-transcribe. OpenAI will remove the latter three from its API on 02/26/2027; a model already selected in the options will continue to be used until then.
Available Mistral models: Mistral Small (default), Mistral Medium, Mistral Large, Codestral, and Ministral 3 (3B/8B/14B). Transcription: Voxtral Mini. Responses from reasoning Mistral models (thinking blocks) are supported.
Available Gemini models: Gemini 3.8 Flash (default, with reduced thinking to respond within one to two seconds) and Gemini 3.5 Flash-Lite (the fastest). Transcription: Gemini 3.5 Transcribe, through Google's Interactions API, without storing dictations in that API's history (store: false, see docs/providers/gemini.md).
Settings that used gpt-4.1-nano (OpenAI API shutdown on 10/23/2026) or gpt-4o automatically switch to gpt-5.6-luna and gpt-4.1 during the update.
Independent provider selection for transcription and translation/rephrasing.
Compatibility with LiteLLM Proxy through the Custom provider to connect to alternative models.
Comprehensive internationalization management through language files (_locales), providing a multilingual interface and voice support.
🌐 Supported Languages
Here is the list of languages supported by Babel Fish AI, with links to demonstration videos:

Arabic
German
English
Spanish
French
Hindi
Italian
Japanese
Korean
Dutch
Polish
Portuguese
Romanian
Swedish
Chinese