Supertonic FastAPI - High Performance OpenAI-Compatible TTS API

Supertonic FastAPI is a production-ready, lightning-fast, and OpenAI-compatible text-to-speech (TTS) API powered by Supertonic. It is designed for high-concurrency environments, providing a seamless drop-in replacement for OpenAI's speech synthesis services while offering superior performance and flexibility.

Key Features & SEO Benefits

• OpenAI API Compatibility: Full support for OpenAI's TTS endpoints and data formats. Use existing SDKs without modification.
• Advanced Auto-Chunking: Automatically handles long text inputs, ensuring smooth and consistent audio generation for long-form content.
• Multilingual & Multi-Voice: Supports a wide range of professional voices (F1-F5, M1-M5) mapped to standard OpenAI voice names (alloy, echo, fable, onyx, nova, shimmer).
• Real-Time Streaming: Implements chunked transfer encoding, allowing users to play audio while it's still being generated (LLM-ready).
• GPU & CPU Hardware Acceleration: Optimized for NVIDIA CUDA (GPU) and Apple Silicon CoreML (Mac) using ONNX Runtime for near-instant inference.
• Dockerized for Scale: Ready for containerized deployment with Nginx load balancing and multi-process support.
• Professional Speech Synthesis: High-quality, natural-sounding AI voices suitable for podcasts, narrations, and assistants.

Quick Start (Local)

1. Setup:

   ./scripts/setup.sh

2. Run:

   ./scripts/start.sh
   # Server listens on http://0.0.0.0:8800

Production Deployment (Docker)

The easiest way to run in production is using Docker Compose, which includes an Nginx load balancer.

Using Docker Compose

1. Start:

   docker compose up -d

2. Scale:

   # Scale to 3 worker instances
   docker compose up -d --scale api=3

Individual Image

1. Build:

   docker build -t supertonic-tts .

2. Run:

   # CPU / Mac (Recommended for Docker on Mac)
   docker run -p 8800:8800 supertonic-tts
   
   # NVIDIA GPU (Requires nvidia-container-toolkit)
   docker run --gpus all -p 8800:8800 supertonic-tts

Configuration (.env)

Variable	Default	Description
PORT	8800	Port to listen on
FORCE_PROVIDERS	auto	Force specific ORT provider: cuda, coreml, cpu, or auto
MODEL_THREADS	0	Intra-op threads (0=auto)
MAX_WORKERS	4	Thread pool workers for concurrent requests

API Usage

POST /v1/audio/speech

{
  "model": "tts-1",
  "input": "Hello, this is a test.",
  "voice": "alloy",
  "stream": true
}

Get Voices: /voices Health Check: /health