Media
5 items
Description
Real-time audio transcription - 100% local AI TabVoice is your free AI transcription assistant for the browser. It turns audio from any tab; YouTube videos, podcasts, webinars, online courses, and meetings into live text, right as it plays. Everything runs on your own computer. No account. No subscription. No API keys. 100% free. WHY TABVOICE? Most transcription tools charge monthly fees or send your audio to remote servers creating risks of data breaches, third-party AI training, and compliance headaches under GDPR, LGPD, and other privacy laws. TabVoice takes a different approach: it runs OpenAI Whisper directly in your browser using WebGPU or WASM. Your audio is captured from the tab, transcribed locally, and stored in your browser. Nothing is uploaded. And unlike "free" tools that lock basic features behind a paywall, TabVoice is completely free: every feature, for everyone. KEY FEATURES • Real-time transcription from any browser tab • 100% free: no license, no account, no paywall • Automatic hardware detection recommends the best AI model for your machine • Choose your model: from lightning-fast (tiny) to high-accuracy (turbo) • 100% local processing: no audio ever sent to the cloud • Works offline after the first model download • Floating captions overlay on any webpage • Unlimited saved sessions with search, tags, and filters • Export transcriptions as .txt • Privacy by design — your audio never leave your device HOW IT WORKS 1. Open any tab with audio. 2. Click Start Capture in the TabVoice side panel. 3. The AI model (Whisper) transcribes the audio locally in your browser. 4. Read the transcript in the side panel or enable the floating overlay. 5. Stop when done. Your session is saved, searchable, and exportable. PRIVACY BY DESIGN TabVoice never uploads your audio or transcripts — transcription happens entirely inside your browser, and you can verify this in Chrome Developer Tools: no outgoing audio requests during capture. The only things that ever leave your device are the one-time AI model download (from Hugging Face, cached locally from then on) and anonymous usage statistics that help us improve the product. No transcript content, no URLs, no audio. Ever. LIMITATIONS • Currently optimized for English audio. Support for more languages is coming. • The first use requires downloading an AI model (~40 MB). • Speaker labels (who said what) are not yet supported. • Word-level timestamp syncing with video is not yet available. • Requires a modern browser with WebGPU or WASM support. TabVoice is free — today and going forward. Download it and start transcribing.