NeuTTS OpenAI API

A FastAPI-based text-to-speech service using NeuTTS Air with full OpenAI TTS streaming API compatibility.

🎯 OpenAI TTS Streaming API Compatible

This service is fully compatible with the OpenAI TTS streaming API. Set up as shown below, then use with:

• Pipecat - Replace base URL with localhost:8000/v1/
• LiveKit - Replace base URL with localhost:8000/v1/
• OpenWebUI - Replace base URL with localhost:8000/v1/

Features

• ✅ OpenAI API compatible endpoints - Drop-in replacement for OpenAI TTS
• ✅ Streaming audio generation - Real-time audio streaming
• ✅ Multiple audio formats - MP3, WAV, PCM, Opus, AAC, FLAC
• ✅ GPU-accelerated inference - CUDA 12.8 support
• ✅ Multiple voice support - Dave, Jo voices
• ✅ Low-latency generation - ~350ms median latency

Quick Start

Option 1: Manual Setup

1. Clone the repository

   git clone https://github.com/Edward-Zion-Saji/neutts-openai-api.git
   cd neutts-openai-api
   

2. Create and activate virtual environment

   python -m venv venv
   source venv/bin/activate
   

3. Install dependencies

   pip install -r requirements.txt
   

4. Run the application

   python openai.py
   

The API will be available at http://localhost:8000

Option 2: Docker with GPU Support

1. Clone the repository

   git clone https://github.com/Edward-Zion-Saji/neutts-openai-api.git
   cd neutts-openai-api
   

2. Build the Docker image

   docker build -t neutts-openai-api .
   

3. Run with GPU support

   Option A: Docker run

   docker run --gpus all -p 8000:8000 neutts-openai-api
   

   Option B: Docker Compose (recommended)

   docker-compose up --build
   

API Endpoints

OpenAI Compatible Endpoint

POST /v1/audio/speech

Request:

{
  "input": "Hello, this is a test",
  "voice": "coral",
  "response_format": "mp3"
}

Custom Endpoints

• POST /synthesize - Basic TTS synthesis
• POST /synthesize-with-timing - TTS with timing information
• GET /health - Health check
• GET / - Web interface

Supported Formats

• MP3 - audio/mpeg
• WAV - audio/wav
• PCM - audio/pcm
• Opus - audio/opus
• AAC - audio/aac
• FLAC - audio/flac

Voice Options

• coral - Maps to Dave voice
• dave - Dave voice

Requirements

• Python 3.10+
• CUDA-compatible GPU (for GPU acceleration)
• Linux (recommended for Docker)

Docker Requirements

• Docker with GPU support
• NVIDIA Container Toolkit
• CUDA drivers

Docker Setup (Linux)

1. Install NVIDIA Container Toolkit

   distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
   curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
   curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
   
   sudo apt-get update && sudo apt-get install -y nvidia-docker2
   sudo systemctl restart docker
   

2. Verify GPU support

   docker run --rm --gpus all nvidia/cuda:12.8-base-ubuntu22.04 nvidia-smi
   

Integration Examples

Pipecat Integration

from pipecat.pipeline.pipeline import Pipeline
from pipecat.services.openai import OpenAILLMService
from pipecat.services.elevenlabs import ElevenLabsTTSService

# Replace OpenAI TTS with NeuTTS
pipeline = Pipeline([
    OpenAILLMService(api_key="your-key", model="gpt-4"),
    ElevenLabsTTSService(api_key="your-key", voice_id="your-voice")
])

# Change to use local NeuTTS service
pipeline = Pipeline([
    OpenAILLMService(api_key="your-key", model="gpt-4"),
    ElevenLabsTTSService(api_key="dummy", voice_id="coral", base_url="http://localhost:8000/v1/")
])

LiveKit Integration

from livekit.agents import JobContext, WorkerOptions, cli
from livekit.agents.voice_assistant import VoiceAssistant
from livekit.plugins import openai

# Configure OpenAI TTS to use local NeuTTS
openai.TTS_BASE_URL = "http://localhost:8000/v1/"

OpenWebUI Integration

# In your OpenWebUI configuration
openai:
  base_url: "http://localhost:8000/v1/"
  api_key: "dummy"  # Not used by NeuTTS

Documentation

Documentation available at /docs when the service is running.