Kokoro-FastAPI-GPU

Kokoro-FastAPI-GPU

Docker app from pureelectricity's Repository

Overview

Kokoro-FastAPI is a Dockerized FastAPI wrapper for the Kokoro-82M text-to-speech model, optimized for speed and quality.

Features:
- High-quality Japanese and English text-to-speech.
- Accelerated inference using NVIDIA GPUs.
- OpenAI-compatible speech endpoint.

Home Assistant Integration:
This container is fully compatible with the OpenAI TTS integration in Home Assistant. You can generate local, OpenAI-level quality voice audio using your own GPU.
1. Install the OpenAI TTS integration (via HACS).
2. Set the "API Key" to any random text (e.g., "123").
3. Set the "Target URL" to: http://[IP]:8880/v1

Requirements:
- Requires the Nvidia Driver plugin.
- This container runs as internal user 1001. Ensure your "Temp Files" output directory is writable by this user.
- The WebUI is located at /web.

Requirements

Nvidia Driver

Runtime arguments

Web UI
http://[IP]:[PORT:8880]/web
Network
bridge
Shell
bash
Privileged
false
Extra Params
--gpus all --restart=unless-stopped

Template configuration

WebUI PortPorttcp

Port for the FastAPI server.

Target
8880
Default
8880
Value
8880
Temp Files (Audio Output)Pathrw

Directory where generated audio files are saved. MUST be writable by UID 1001.

Target
/app/api/temp_files
Default
/mnt/user/appdata/kokoro-tts/temp
NVIDIA_VISIBLE_DEVICESVariable

GPU UUID to pass to the container (or 'all').

Default
all
Value
all
NVIDIA_DRIVER_CAPABILITIESVariable

Nvidia driver capabilities.

Default
all
Value
all
Use GPUVariable

Explicitly enable GPU usage.

Target
USE_GPU
Default
true
Value
true
Device TypeVariable

Set inference device (gpu or cpu).

Target
DEVICE
Default
gpu
Value
gpu

Details

Repository
ghcr.io/remsky/kokoro-fastapi-gpu:latest
Last Updated2026-05-31
First Seen2025-04-25

Run Kokoro-FastAPI-GPU on Unraid.

Kokoro-FastAPI-GPU is listed in Community Apps for Unraid OS. Explore Unraid to build a flexible home server, NAS, or homelab.