Svelte Voice Notes Transcription

A modern web application built with SvelteKit that demonstrates two different approaches to speech-to-text transcription: browser-native Speech Recognition API and OpenAI's Whisper API. This app allows users to record, transcribe, and manage voice notes using either transcription method.

Demo

Watch the demo video

Features

• Dual transcription methods:
  • Browser's native Speech Recognition API for real-time transcription
  • OpenAI's Whisper API for high-accuracy transcription
• Voice recording controls (start, stop, pause, resume)
• Note management system (create, save, load, delete)
• Real-time transcription display
• Persistent storage of notes using localStorage

Architecture

Core Components

1. Speech Handlers:

   • SpeechHandler: Manages browser-native speech recognition
   • SpeechHandlerOpenAi: Handles Whisper API integration

2. State:

   • VoiceNotesHandler: Manages note storage and retrieval
   • Svelte context API for state sharing

3. UI Components:

   • Recorder: Controls for voice recording
   • CreateDialog: Note creation interface
   • LoadNoteDialog: Note loading interface

Implementation Details

Browser Speech Recognition

// Uses the Web Speech API
const SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
this.recognition = new SpeechRecognition();
this.recognition.continuous = true;
this.recognition.interimResults = true;

OpenAI Whisper Integration

// Handles audio chunks and sends to Whisper API
async transcribeAudio(audioBlob: Blob): Promise<string> {
    const file = new File([audioBlob], 'recording.webm', { type: MIME_TYPE });
    const formData = new FormData();
    formData.append('file', file);

    const response = await fetch('/api/transcribe', {
        method: 'POST',
        body: formData
    });
}

Key Features Implementation

Recording Controls

The app provides a comprehensive set of recording controls:

• Start/Stop recording
• Pause/Resume recording
• Real-time transcription display
• Error handling and user feedback

Note Management

Notes are managed through the VoiceNotesHandler class:

• Create new notes with titles and transcriptions
• Update existing notes
• Delete notes
• Load and display saved notes
• Persistent storage using localStorage

Usage

1. Starting a Recording:

   • Click the "Start Recording" button
   • Grant microphone permissions when prompted
   • Speak into your microphone

2. Managing Recordings:

   • Use the pause/resume button to temporarily stop recording
   • Click "Stop Recording" to finish
   • Save the transcription as a note

3. Managing Notes:

   • Create new notes with the "+" button
   • Load existing notes using the load dialog
   • Edit transcriptions directly in the textarea
   • Copy transcriptions to clipboard
   • Delete unwanted notes

Setup

1. Clone the repository
2. Install dependencies:

   pnpm install

3. Set up environment variables:

   OPENAI_API_KEY=your_api_key_here

4. Start the development server:

   pnpm dev

Configuration

The app includes configurable parameters:

• MIN_CHUNK_SIZE: Minimum size for audio chunks
• DEFAULT_INTERVAL: Default recording interval
• DEFAULT_CONFIDENCE: Default confidence threshold for transcription