Piper Tutorial: Deploy Offline TTS with Docker in 10 Minutes

Piper Tutorial: From Install to First Voice

In this tutorial you'll get Piper up and running with Docker, download a neural voice, and generate your first speech file — all on your own hardware, with zero cloud dependency.

Step 1: Launch the container

Create a docker-compose.yml and start the official community image:

services:
  piper:
    image: linuxserver/piper:latest
    restart: unless-stopped
    ports:
      - 8080:8080
    volumes:
      - ./data/piper:/data

Run docker compose up -d. Piper's CLI is now available inside the container at /usr/local/bin/piper.

Step 2: Download a voice

Voices are small ONNX models with JSON configs. Grab an English voice:

docker compose exec piper sh -c \
  "wget -q https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_GB/amy/medium/en_GB-amy-medium.onnx \
  && wget -q https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_GB/amy/medium/en_GB-amy-medium.onnx.json \
  && mv en_GB-amy-medium* /data/"

Step 3: Generate speech

docker compose exec piper sh -c \
  "echo 'Welcome to Piper, fast local speech synthesis.' | piper \
     --model /data/en_GB-amy-medium.onnx \
     --output_file /data/welcome.wav"

You now have welcome.wav on your host in ./data/piper/ — generated locally in real time.

Step 4: Serve it over HTTP

Piper includes an HTTP server, so any app on your network can request speech:

docker compose exec piper sh -c \
  "python3 -m piper.http_server --model /data/en_GB-amy-medium.onnx"

That's it — under ten minutes from zero to local neural speech. Piper proves that high-quality TTS no longer needs the cloud.