Voice Notes Assistant

Voice Notes Assistant turns spoken audio — meetings, lectures, interviews, or personal voice memos — into clean, readable text in seconds. Record directly in the browser or upload an audio file, and Speechmatics' speech recognition engine handles the heavy lifting of turning speech into an accurate transcript, in either English or Urdu. What sets it apart is how it handles real conversations with multiple speakers. Rather than relying solely on automatic speaker detection, users can record or upload separate clips for each speaker, label them (e.g. "Speaker 1," "Interviewer," a name), and add them one by one to a running queue — nothing is lost between additions. When ready, the app transcribes every clip and stitches them into a single, clearly labeled transcript. Speechmatics' built-in diarization is also available for cases where multiple people are already talking in one recording. Once transcribed, a lightweight, fully local TextRank summarization algorithm — no paid summarization API required — extracts the most important sentences into a concise summary, with an adjustable length from 1–6 sentences. Both the transcript and summary are shown in an in-app preview first, so users can review the content before choosing to download it as a .txt file, rather than a file landing on their machine unreviewed. The interface is designed to feel like a genuinely bilingual tool rather than an English app with Urdu bolted on: Urdu output renders in real Nastaliq script with correct right-to-left layout, not transliterated Latin text. The API key is never exposed in the browser — it's read securely from a server-side environment variable — and the whole experience is wrapped in a playful, mobile-app-inspired dark UI with a mint accent, flag-based language indicators, and rounded, tactile control