🐧 Installing Piper TTS on Linux — Step-by-Step Guide

[TacBotDEV]

️ What is Piper TTS?

Piper is a fast, neural text-to-speech engine that runs 100% offline and locally. It uses espeak-ng for phonemization and works on any Linux distribution.

It supports 40+ languages, including Brazilian Portuguese (pt_BR), with high-quality voices.

---

Installing on Linux

1️⃣ Install Python 3.9+ and pip

Debian / Ubuntu / Linux Mint:

sudo apt update
sudo apt install python3 python3-pip python3-venv -y

Fedora / RHEL / CentOS:

sudo dnf install python3 python3-pip

Arch Linux / Manjaro:

sudo pacman -S python python-pip

openSUSE:

sudo zypper install python3 python3-pip

---

2️⃣ Install Piper

pip install piper-tts

Recommendation: Use a virtual environment to avoid conflicts:

python3 -m venv piper-env
source piper-env/bin/activate
pip install piper-tts

---

3️⃣ Download a Voice

python3 -m piper.download_voices

To download a specific voice:

# Brazilian Portuguese
python3 -m piper.download_voices pt_BR-edresson-medium

# American English
python3 -m piper.download_voices en_US-lessac-medium

Voices are available at huggingface.co/rhasspy/piper-voices.

---

4️⃣ Test the Voice

python3 -m piper -m pt_BR-edresson-medium -f test.wav -- "Hello! This is a test message from Piper TTS on Linux."

To hear it directly (requires ffplay):

# Install ffplay
sudo apt install ffmpeg   # Debian/Ubuntu
sudo dnf install ffmpeg   # Fedora

# Test
python3 -m piper -m pt_BR-edresson-medium -- "Hello, world!"

---

5️⃣ Using the Python API

import wave
from piper import PiperVoice, SynthesisConfig

voice = PiperVoice.load("pt_BR-edresson-medium.onnx")

with wave.open("output.wav", "wb") as wav_file:
    voice.synthesize_wav("Welcome to Piper TTS on Linux!", wav_file)

syn_config = SynthesisConfig(
    volume=1.0,
    length_scale=1.1,
    noise_scale=0.9,
    normalize_audio=True
)
with wave.open("adjusted.wav", "wb") as w:
    voice.synthesize_wav("Custom phrase...", w, syn_config=syn_config)

---

6️⃣ HTTP Server (for continuous use)

To avoid reloading the model for each phrase, start the server:

python3 -m piper.http_server

Access at: http://localhost:5000

Example with curl:

curl -X POST http://localhost:5000/synthesize \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello Piper server!", "voice": "pt_BR-edresson-medium"}' \
  --output audio.wav

---

7️⃣ GPU Acceleration (optional)

With an NVIDIA GPU:

pip install onnxruntime-gpu

# In code
voice = PiperVoice.load("voice.onnx", use_cuda=True)

---

Linux Tips

• Home Assistant Integration: Piper is the default TTS engine in Home Assistant OS
• Automation: Use piper in shell scripts for voice notifications
• Batch processing: cat phrases.txt | python3 -m piper -m voice -f output.wav
• Native binaries (without Python): Download from github.com/rhasspy/piper/releases for amd64/arm64/armv7

---

Useful links:

• Voice samples
• Source code
• Python API docs

Questions? Ask here!