README.md

---
title: VoiceScript
emoji: 🎙️
colorFrom: indigo
colorTo: purple
sdk: docker
pinned: true
license: mit
short_description: AI Speech Recognition — OpenAI Whisper + Facebook Demucs
---

# 🎙️ VoiceScript — AI Speech Recognition

Upload any audio or record live — VoiceScript transcribes it instantly using a professional 3-stage AI pipeline.

## What it does

- 🎤 **Upload or record** — WAV, MP3, FLAC, OGG, WebM supported
- 🧠 **Whisper medium** — OpenAI's powerful speech-to-text model
- 🎵 **Demucs vocal isolation** — strips background music before transcribing
- 🌍 **Any language → English** — Whisper auto-detects 99 languages
- 🔤 **Translate transcript** — to 55+ languages via Google Translate
- ⏱️ **Timestamps** — every sentence tagged with [MM:SS]
- 📄 **Export** — TXT, SRT subtitles, or PDF

## How to use

1. Upload an audio file or click **Start Recording**
2. Choose mode: **Transcribe** or **Any Language → English**
3. Click **Transcribe File** and wait for the pipeline to process
4. Use the Translate panel, Timestamps toggle, or Export buttons

## Processing time

Audio is processed through Demucs vocal isolation + Whisper medium on CPU. Expect roughly 1 minute per 5 minutes of audio.

## Built with

Python · Flask · OpenAI Whisper · Facebook Demucs · pydub · ffmpeg · deep-translator

---

Built by [Tushar Tamrakar](https://github.com/TUSHARTAMRAKAR) · [GitHub Repo](https://github.com/TUSHARTAMRAKAR/VoiceScript)