ViewSym Audio Assistant is a Chrome extension that captures live audio from any browser tab, transcribes it in real-time using Whisper tiny, and provides an AI chat interface (Flan-T5 base) to ask questions about the transcription. Built with Manifest V3.
- Real-Time Transcription — captures tab audio and transcribes with Whisper tiny in 5s chunks (configurable via settings)
- AI Chat — ask questions, request summaries, or extract action items from the transcript
- Adjustable chunk size — control latency vs. accuracy tradeoff (1s–15s)
- Settings dropdown — full-width panel with model status and export
- Clone or download this repository.
- Open Chrome and navigate to
chrome://extensions/. - Enable Developer mode.
- Click Load unpacked and select the
viewsym-extdirectory.
- Open a tab with audio (YouTube, meeting, podcast, etc.).
- Open the ViewSym side panel (click the extension icon or use the keyboard shortcut).
- Click Start in the panel header to begin capture.
- View the live transcript and ask questions in the Ask Assistant section.
- Click Stop when done.
Click the gear icon in the header to open the settings dropdown:
- Model status — shows Whisper tiny (transcription) and Flan-T5 base (chat) load state
- Chunk Size — slider (1000–15000ms) to control how often audio is transcribed; larger = more accurate, smaller = lower latency
- Download transcript — exports the full transcript as a
.txtfile
- Service Worker (
background.js) — ensures offscreen document is alive, relays messages - Offscreen Document (
offscreen.html/offscreen.js) — captures tab audio viaMediaRecorder, transcribes chunks with Whisper tiny via Hugging Face Transformers.js - Side Panel (
sidepanel.html/sidepanel.js) — UI with transcript view, chat interface, and settings
sidePanel— render UI alongside web contentstorage— persist chunk size preferencetabCapture— capture audio from the active tabactiveTab— interact with the current taboffscreen— run a hidden document for audio processing
MIT