README.md

metadata

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 · GitHub Repo