Kokoro FastAPI is a Dockerized, OpenAI-compatible API server for Kokoro-82M, the compact open text-to-speech model — it lets you generate hours of high-quality speech in minutes from any app that can call the OpenAI speech endpoint. It supports multiple languages and speakers, voice clone-tuning, caption timestamps and SSML, and runs on CPU, AMD and NVIDIA GPUs through PyTorch. The project has about 5,600 stars on GitHub and is actively maintained, with updates in 2026.
Kokoro-82M punches well above its size for speech quality, but using it directly means writing model code. Kokoro FastAPI wraps it in a production-friendly FastAPI service: point your existing OpenAI-compatible client at the endpoint and you get speech synthesis with no code changes, which is why it integrates neatly with tools like Open WebUI and SillyTavern. Language support covers English (US and GB), Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese and Mandarin Chinese, with custom voicepacks available through the companion clone-tuner. An optional read-along web UI highlights text as it is spoken, long-form generation handles books and scripts, and features such as phoneme-level control, normalization controls and caption timestamps make it suitable for audiobooks, accessibility and content pipelines, all self-hosted via Docker.
Kokoro FastAPI is free and open source under the Apache-2.0 license, as is the Kokoro-82M model it serves. You only pay for the hardware or server you host it on.
Kokoro FastAPI is for developers who want self-hosted text-to-speech behind a familiar API — for voice features in apps, audiobook and video narration pipelines, accessibility tooling, or local AI companions — without per-character cloud TTS bills.