tts-server-proxy

Kokoro-FastAPI — ★★☆ Medium

Repo: https://github.com/remsky/Kokoro-FastAPI · License: Apache-2.0 GUI: optional Web UI · Cloning: ❌ preset voices only · Italian: ✅ preset voices Default port: 8880

FastAPI wrapper around Kokoro-82M: an OpenAI-compatible speech endpoint with multi-language preset voices (English US/GB, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, Mandarin), ElevenLabs-compatible websockets, and an optional integrated Web UI. Very fast even on CPU.

Install (Windows, native — no Docker needed)

  1. Install uv (single PowerShell command).
  2. Clone the repo and use the direct-run path:

    git clone https://github.com/remsky/Kokoro-FastAPI
    cd Kokoro-FastAPI
    uv run python -m api.main  # native run; models auto-download
    

    (Windows PowerShell helper scripts are also provided upstream; Docker images kokoro-fastapi-cpu / kokoro-fastapi-gpu are the alternative if you prefer containers.)

  3. Optional Web UI at http://localhost:8880/web.

Connect tts-server-proxy

App field Value
Endpoint http://<pc-ip>:8880
API key empty (or the key you configured on the server — Bearer is honored)
model kokoro
voice Italian presets: if_sala (female) or im_nicola (male)
lang it

Gotchas