Kokoro TTS API

FastAPI wrapper for Kokoro-82M text-to-speech model, providing an OpenAI-compatible endpoint with:

• NVIDIA GPU accelerated inference (or CPU) option
• automatic chunking/stitching for long texts
• very fast generation time (~35-49x RTF)

Quick Start

1. Install prerequisites:

   • Install Docker Desktop
   • Install Git (or download and extract zip)

2. Clone and start the service:

# Clone repository
git clone https://github.com/remsky/Kokoro-FastAPI.git
cd Kokoro-FastAPI

# For GPU acceleration (requires NVIDIA GPU):
docker compose up --build

# For CPU-only deployment (~10x slower, but doesn't require an NVIDIA GPU):
docker compose -f docker-compose.cpu.yml up --build

Test all voices (from another terminal):

python examples/test_all_voices.py

Test OpenAI compatibility:

python examples/test_openai_tts.py

OpenAI-Compatible API

List available voices:

import requests

response = requests.get("http://localhost:8880/audio/voices")
voices = response.json()["voices"]

Generate speech:

import requests

response = requests.post(
    "http://localhost:8880/audio/speech",
    json={
        "model": "kokoro",  # Not used but required for compatibility
        "input": "Hello world!",
        "voice": "af_bella",
        "response_format": "mp3",  # Supported: mp3, wav, opus, flac
        "speed": 1.0
    }
)

# Save audio
with open("output.mp3", "wb") as f:
    f.write(response.content)

Using OpenAI's Python library:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8880", api_key="not-needed")

response = client.audio.speech.create(
    model="kokoro",  # Not used but required for compatibility, also accepts library defaults
    voice="af_bella",
    input="Hello world!",
    response_format="mp3"
)

response.stream_to_file("output.mp3")

Performance Benchmarks

Benchmarking was performed on generation via the local API using text lengths up to feature-length books (~1.5 hours output), measuring processing time and realtime factor. Tests were run on:

• Windows 11 Home w/ WSL2
• NVIDIA 4060Ti 16gb GPU @ CUDA 12.1
• 11th Gen i7-11700 @ 2.5GHz
• 64gb RAM
• WAV native output
• H.G. Wells - The Time Machine (full text)

Key Performance Metrics:

• Realtime Factor: Ranges between 35-49x (generation time to output audio length)
• Average Processing Rate: 137.67 tokens/second (cl100k_base)

Features

• OpenAI-compatible API endpoints
• GPU-accelerated inference (if desired)
• Multiple audio formats: mp3, wav, opus, flac, (aac & pcm not implemented)
• Natural Boundary Detection:
  • Automatically splits and stitches at sentence boundaries to reduce artifacts and maintain performacne

Note: CPU Inference is currently a very basic implementation, and not heavily tested

Model

This API uses the Kokoro-82M model from HuggingFace.

Visit the model page for more details about training, architecture, and capabilities. I have no affiliation with any of their work, and produced this wrapper for ease of use and personal projects.

License

This project is licensed under the Apache License 2.0 - see below for details:

• The Kokoro model weights are licensed under Apache 2.0 (see model page)
• The FastAPI wrapper code in this repository is licensed under Apache 2.0 to match
• The inference code adapted from StyleTTS2 is MIT licensed

The full Apache 2.0 license text can be found at: https://www.apache.org/licenses/LICENSE-2.0

Sample

https://user-images.githubusercontent.com/338912d2-90f3-41fb-bca0-5db7b4e02287.mp4