remsky/Kokoro-FastAPI
DOCKER WRAPPER KOKORO-82M TTS WITH OPENAI API
Docker image with FastAPI wrapper over Kokoro-82M speech synthesis model. Provides OpenAI-compatible API, websockets for voice agents, and web UI for audio generation.
// KEY FEATURES
- OpenAI-compatible API and websockets for LiveKit/Pipecat
- Multilingual: 8 languages including English, Chinese, Japanese
- Multi-speaker, SSML, word timestamps, voice cloning
- Prebuilt images for CPU, NVIDIA CUDA, AMD ROCm, Apple MPS
// DEEP DIVE: HOW IT WORKS
Project packages Kokoro-82M into FastAPI with OpenAI-compatible API (/v1/audio/speech, realtime). Weights load with sha256 check of pinned HF commit; on mismatch server won't start. Images for linux/amd64/arm64: CPU, NVIDIA CUDA (cu126/128/129), AMD ROCm (exp.); on Apple Silicon — native via MPS.
ElevenLabs/Cartesia websockets for LiveKit/Pipecat, Inno Clone-Tuner for custom voices, optional WebUI with read-along. SSML, inline multi-speaker with weights, phonemes, word/chunk timestamps. 8 languages: en-US/GB, es, fr, hi, it, ja, pt-BR, zh.
Requires Docker; for GPU — NVIDIA Container Toolkit or ROCm. CPU image on any x86_64/arm64. Model ~300 MB cached. No auth — deploy in trusted network or behind proxy. In prod pin image tag, not :latest.
// USE CASES
Best for: Developers of voice agents and TTS services for local deployment
- Deploy local TTS backend for voice assistant on LiveKit
- Generate audiobook with multiple voices and timestamps for sync
- Create voice clone via Clone-Tuner and use via OpenAI API
// QUICK START
docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:latest
docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest
docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest-cu128
docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest
docker run --device=/dev/kfd --device=/dev/dri -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-rocm:latest
./start-gpu_mac.shCommands are taken from the repository README — check the current version on GitHub before running.
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram