A TypeScript project that explores OpenAI.fm's text-to-speech capabilities through API reverse engineering. Turn any length of text into natural-sounding speech with parallel processing and smart chunking.

Educational Notice: This is a learning project created to understand API interactions and audio processing. Not intended for production use. Please respect API terms of service.

• Unlimited Text Length: Handles any text size through smart chunking
• 3x Faster Processing: Parallel chunk processing (3 simultaneous chunks)
• Clean Audio: 250ms pauses between chunks for natural transitions
• Multiple Voices: 7 different voice options
• Speaking Styles: Various vibes from professional to storytelling

# Clone the repository
git clone https://github.com/springmusk026/OpenFM-API.git

# Change directory
cd OpenFM-API

# Install dependencies
npm install

# Run the example
npm start

• shimmer
• echo
• fable
• onyx
• nova
• breeze
• ember

• professional
• empathetic
• storytelling
• newscast
• motivational
• customer_service
• meditation

// Simple and efficient chunking at sentence boundaries
public chunkText(text: string): string[] {
    // Find last period before 1000 char limit
    const segment = text.slice(0, this.maxChunkSize);
    const lastPeriod = Math.max(
        segment.lastIndexOf('. '),
        segment.lastIndexOf('! '),
        segment.lastIndexOf('? ')
    );
}

// Clean chunk transitions with 250ms pauses
public static async mergeAudioBuffers(
    buffers: ArrayBuffer[]
): Promise<ArrayBuffer> {
    // Add 250ms silence between chunks
    const pauseBuffer = createSilentBuffer(250);
    // Merge with pauses
}

// Process 3 chunks simultaneously
public async generateAudio(text: string): Promise<ArrayBuffer> {
    const chunks = this.textChunker.chunkText(text);
    const batchSize = 3;
    for (let i = 0; i < chunks.length; i += batchSize) {
        const batch = chunks.slice(i, i + batchSize);
        await Promise.all(batch.map(chunk => this.processChunk(chunk)));
    }
}

src/
├── services/
│   ├── OpenFMService.ts     # API interaction & parallel processing
│   └── TextChunkerService.ts # Smart text chunking
├── utils/
│   └── AudioUtils.ts        # Audio merging & processing
├── types/
│   └── index.ts            # TypeScript definitions
└── index.ts                # Example usage

1. Simplified Text Chunking

   • Just finds last sentence boundary before limit
   • No complex preprocessing
   • Fast and reliable

2. Better Audio Transitions

   • 250ms silence between chunks
   • No more crossfading complexity
   • Cleaner speech output

3. Parallel Processing

   • 3 chunks processed simultaneously
   • Maintains correct order
   • Significant speed improvement

This is an educational project - improvements welcome! Areas to enhance:

• Better error handling
• More voice options
• Improved audio processing
• Web interface
• Documentation

• API reverse engineering techniques
• Audio processing in TypeScript
• Parallel processing patterns
• Clean code practices

MIT License - Free to use for educational purposes.

This is a learning project for understanding API interactions and audio processing. Not intended for production use.