VoiceCipher 🎙️

Local Speech transcription using Tiny Whisper model

Welcome to VoiceCipher, a static web application that uses transformer.js to run the Tiny Whisper model directly through JS on the user's local machine to transcribe audio to English only.

Features 🌟

• User-Friendly Interface: Clean and intuitive design for a seamless experience. 🎨
• Flexible Input Options: Transcribe from URLs, file uploads, or direct recordings. 🔗📁🎤
• Speech-to-Text Transcription: Convert your voice recordings into text using state-of-the-art ML models. 🧠

Getting Started 🚀

To get started with VoiceCipher, follow these steps:

1. Clone the Repository

   git clone https://github.com/bharathajjarapu/voicecipher.git
   

2. Navigate to the Project Directory

   cd voicecipher
   

3. Go Live or Run the HTML File Simply open index.html in your browser to launch the application. 🌐

Usage 🛠️

1. Choose Your Input Method:

   • From URL: Enter the URL of the audio file. 🌐
   • From File: Upload an audio file from your computer. 📁
   • Record: Record audio directly through your microphone. 🎤

2. Transcribe: Click the "Transcribe" button to start converting your audio into text. ✏️

3. View Results: The transcription will appear in the text area below the buttons. Note that it is currently limited to 20 seconds for most devices; improvements are planned. ⏱️

How It Works 🔍

VoiceCipher uses transformers.js to run the Whisper model for speech-to-text conversion. Here's a brief explanation:

• Transformer.js: This library is used to interface with the Whisper model, enabling speech recognition directly in the browser. 📊
• Static Execution: Unlike many ML models that rely on WebAssembly for execution, Whisper runs statically without WebAssembly. This approach simplifies deployment and ensures compatibility across different environments. ⚙️

Dependencies 📦

• Bootstrap for styling. 🎨
• Xenova Transformers for speech recognition. 🧠

Future Ideas 🚧

1. Adding support for audio clips longer than 20 seconds for all devices.
2. Adding a multilingual model to transcribe various languages, like Telugu. 🌍

License 📝

This project is licensed under the MIT License. See the LICENSE file for details.

Contact 📫

For any questions or contributions, please contact [email protected].

Made using Transformer.js 🚀