Deploy and Host Kokoro-FastAPI on Railway

Kokoro-FastAPI is a Dockerized FastAPI wrapper for the Kokoro-82M text-to-speech model. It provides OpenAI-compatible endpoints for multi-language speech synthesis with support for multiple audio formats, streaming, and voice combinations.

Resource Requirements — Read Before Deploying

This template runs the PyTorch build of Kokoro and needs at least 1.5 GB of RAM. Measured on the current image: ~1.05 GB resident once the model has loaded, ~1.15 GB after serving a request, and about 20 seconds of cold start before the API answers.

That means it requires Railway's Hobby plan or above. On the Trial (1 GB) and Free (0.5 GB) plans the container is OOM-killed during model load with exit status 137 and never serves traffic. If you are on one of those plans, deploy Kokoro TTS Lite instead — same OpenAI-compatible /v1/audio/speech, ONNX runtime, ~375 MB resident, runs on the Free plan.

A healthcheck on /health is configured, so Railway waits for the model to finish loading before routing traffic to a new deploy.

About Hosting Kokoro-FastAPI

Deploying Kokoro-FastAPI on Railway gives you an OpenAI-compatible text-to-speech API. The service supports multiple output formats (MP3, WAV, Opus, FLAC, M4A, PCM), streaming audio with configurable chunk sizes, and multi-language synthesis including English, Japanese, and Chinese. Advanced features include weighted voice combinations, word-level timestamped captions, and natural boundary detection for automatic sentence-based splitting. A built-in web UI and API documentation are included.

Common Use Cases

Drop-in replacement for OpenAI's text-to-speech API
Generating audio content in multiple languages
Creating custom voices by combining existing voice profiles
Producing captioned audio with word-level timestamps
Streaming audio generation for real-time applications

Dependencies for Kokoro-FastAPI Hosting

Docker container runtime
Kokoro-82M model (Apache 2.0 licensed)

Deployment Dependencies

Implementation Details

OpenAI-compatible client usage:

from openai import OpenAI

client = OpenAI(
    base_url="https://your-app.up.railway.app/v1",
    api_key="your-api-key"
)

response = client.audio.speech.create(
    model="kokoro",
    voice="af_bella",
    input="Hello world!"
)
response.stream_to_file("output.mp3")

Voice combinations with weighted ratios:

response = client.audio.speech.create(
    model="kokoro",
    voice="af_bella(2)+af_sky(1)",  # 67% bella, 33% sky
    input="Hello world!"
)

API Endpoints:

Endpoint	Purpose
`POST /v1/audio/speech`	OpenAI-compatible speech synthesis
`GET /v1/audio/voices`	List available voices
`POST /v1/audio/voices/combine`	Create weighted voice combinations
`POST /dev/captioned_speech`	Generate audio with timestamps

Web Interfaces:

API Documentation: /docs
Web UI: /web

Why Deploy Kokoro-FastAPI on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Kokoro-FastAPI on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.