Audio Transcription & Live Interpreter

Real-time transcription, live translation, and speech-to-speech playback for browser audio. Includes Subtitle TTS for videos.

Audio Transcription is a powerful extension that turns your browser into a real-time interpreter. It captures any audio playing in a tab (transcribing it via Whisper AI) or reads existing video subtitles, translates them live, and reads the results back to you via Text-to-Speech (TTS). 🌟 Designed with privacy and efficiency in mind, it is optimized to run smoothly on low-resource computers and operates as independently from cloud services as possible. Compatible with Linux, Windows, and macOS, this extension acts as a true Live Interpreter for any media stream. ✨ Key Features: • 🎬 Subtitle TTS Mode: Read aloud and translate existing subtitles from YouTube, Twitch, or any HTML5 video without needing a local server. • 🌐 Source Language Control: Rely on smart auto-detection, or manually select the subtitle language for maximum accuracy. • 🗣️ Real-Time Speech-to-Speech: Listen to live translations with a natural, fluid voice that buffers complete sentences for a seamless experience. • 📝 Live Audio Transcription: Fast and accurate transcription from scratch using your local machine's processing power with OpenAI's Whisper AI (WhisperLive server required). • 🤖 Instant Translation: Translate live text using Google Translate (free) or the latest Google Gemini (Flash-Lite) & Gemma 4 AI models. • 🖼️ Flexible UI Modes: View transcripts in a floating overlay or a dedicated Standalone popup window. • 🛡️ Total Privacy: Local audio processing and transparent open-source code. ⚙️ SERVER INSTRUCTIONS & SOURCE CODE: The Subtitle TTS mode works completely out-of-the-box. However, to use the advanced "Live Audio Transcription" feature, you must run the local WhisperLive server on your computer. Get the server scripts and detailed setup instructions at: https://github.com/antor44/Audio-Transcription 🌐 Universal Real-Time Compatibility This extension is designed with an ambitious goal: to provide a universal, real-time interpretation experience across as many languages, websites, media formats, and transcription methods as possible, whether using native captions or live AI transcription through WhisperLive. This flexibility also comes with challenges. Languages can have very different writing and grammatical structures, including spaceless CJK scripts and right-to-left languages such as Arabic and Hebrew. Video players may also change how subtitles are rendered or include hidden formatting characters that can affect how captions are processed. YouTube can be particularly challenging because many subtitles are automatically generated by YouTube rather than manually edited. They may contain transcription errors, missing punctuation, unusual line breaks, or changes in the way captions are rendered over time. The extension does not pre-download subtitle files such as `.vtt` or `.srt`, even when they are available. Instead, it processes captions strictly in real time as they appear on screen, or as they are generated by the WhisperLive server. Because the system needs enough text to form a logical sentence before sending it to translation and Text-to-Speech, a small amount of latency is inherent to the way the extension works. As a result, occasional edge cases may occur on particular websites, languages, or subtitle formats. The extension is continuously refined to improve compatibility and reliability across different environments. ⚖️ LICENSE: This is a free and open-source project distributed under the GNU General Public License v3.0 (GPL-3.0). For more details, visit the GitHub repository. --- 🆕 WHAT'S NEW IN VERSION 3.4.0: *(All improvements in this update apply specifically to Subtitle TTS mode; WhisperLive capture mode remains unchanged)* • ⚡ YouTube Caption Accuracy: Resolved an issue where rapid, short auto-generated caption blocks could cause repetitive word loops or drop dialogue at the very start of a video. • 🛡️ Ad Continuity: Fixed a bug where commercial breaks dropped the active spoken phrase; sentences paused by ads now resume seamlessly once the video continues. • 📜 Persistent Window History: The transcribed and translated text history in the overlay window is now preserved across commercial breaks and manual timeline seeking, maintaining your reading context without being erased. • 🌐 Improved Language Auto-Detection: Subtitle language detection now dynamically resets when switching videos or changing the menu back to "Auto".