OPEN-SOURCE MIGRATION BENCHMARK License: Apache-2.0 ★ 6.4k GitHub Stars

Replace ElevenLabs / Cloud Speech API with Self-Hosted Kokoro-FastAPI / OpenedAI Speech

OpenAI-compatible drop-in local TTS & Whisper STT server that runs fast even on CPU/VPS Stop paying Pro ($99/mo character quota) ($99/mo for a 10-person team) and run Kokoro-FastAPI / OpenedAI Speech on a $4.50/mo Linux VPS with full data sovereignty.

10-Seat Annual Savings

+$1,134 / yr

vs ElevenLabs / Cloud Speech API standard plan

Container RAM & vCPU

1024 MB RAM

2.0 vCPU • Migration: Instant (Change base_url in OpenAI SDK)

Recommended Cloud Box

$4.50/mo VPS

Hetzner CX22 (4GB RAM / 2 vCPU)

1-to-100 Seat Total Cost of Ownership (TCO) Table

Comparing ElevenLabs / Cloud Speech API (Pro ($99/mo character quota)) vs Kokoro-FastAPI / OpenedAI Speech on Hetzner CX22 (4GB RAM / 2 vCPU).

Team Size	ElevenLabs / Cloud Speech API Cost	Kokoro-FastAPI / OpenedAI Speech VPS	Net Annual Savings
1 Seat (Solo)	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year
5 Seats	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year
10 Seats	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year
25 Seats	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year
50 Seats	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year
100 Seats	$99.00/mo ($1,188/yr)	$4.50/mo (CX22)	+$1,134 / year

Production Docker Compose + Caddy Auto-SSL

Copy and run docker compose up -d on any Ubuntu 24.04 server.

Open in Multi-Tool Builder →

# Production Docker Compose: Kokoro-FastAPI / OpenedAI Speech (Replaces ElevenLabs / Cloud Speech API)
services:
  caddy:
    image: caddy:2-alpine
    restart: unless-stopped
    ports:
      - "80:80"
      - "443:443"
    volumes:
      - ./Caddyfile:/etc/caddy/Caddyfile:ro
      - caddy_data:/data
    networks:
      - stack_net

  kokoro-openedai-speech:
    image: ghcr.io/remsky/kokoro-fastapi-cpu:latest
    container_name: stack-kokoro-openedai-speech
    restart: unless-stopped
    environment:
      - "USE_ONNX=true"
    volumes:
      - kokoro_openedai_speech_data:/data
    networks:
      - stack_net

networks:
  stack_net:
    driver: bridge

volumes:
  caddy_data:
  kokoro_openedai_speech_data:

Caddyfile (Automatic HTTPS Reverse Proxy):

# Automatic Let's Encrypt HTTPS for Kokoro-FastAPI / OpenedAI Speech
kokoro-openedai-speech.{$DOMAIN:yourdomain.com} {
  encode zstd gzip
  reverse_proxy kokoro-openedai-speech:8880
}

Why Teams Switch to Kokoro-FastAPI / OpenedAI Speech

✓82M parameter Kokoro TTS sounds studio-grade on CPU
✓100% compatible with `/v1/audio/speech` endpoint
✓Unlimited audio generation for $0

GitHub Repository remsky/Kokoro-FastAPI

Official Website github.com/remsky/Kokoro-FastAPI

Docker Image ghcr.io/remsky/kokoro-fastapi-cpu:latest

Database Needed Built-in / SQLite

1-Click Cloud VPS Hosting

Deploy Kokoro-FastAPI / OpenedAI Speech in 60 Seconds

Sized automatically for Kokoro-FastAPI / OpenedAI Speech (1024MB container RAM + OS overhead).

Hetzner Cloud (CX22)

Includes €20 Free Cloud Credit

$4.50/mo →

DigitalOcean Droplet

Includes $200 60-Day Free Credit

$24.00/mo →

Railway.app Managed Container

Zero server maintenance

$5.00/mo →

PRODUCTION BOILERPLATE ZIP

Get All 16 Pre-Built Production Stacks + Encrypted S3 Backups