edonseki/folsh.ai

Albanian (sq) TTS Voice for Piper

This repository contains a high-quality Albanian (Shqip) Text-to-Speech (TTS) model in .onnx format. This model is optimized for Piper, a fast, local neural text-to-speech engine.

Author

This model was developed/contributed by Edon Sekiraqa.
If you find this model useful, feel free to give it a ⭐️ on Hugging Face!

Model Details

• Language: Albanian (sq)
• Locale: sq_AL
• Format: ONNX
• Library: Piper
• License: Creative Commons (CC)
• Base Model: rhasspy/piper-voices

Installation

To use this model, download both the .onnx file and the corresponding .onnx.json config file. Piper requires both files to be in the same directory to function correctly.

Usage

Command Line Interface (CLI)

After installing Piper, you can generate audio using the following command:

echo "Përshëndetje, ky është një zë i krijuar me inteligjencë artificiale." | \
  ./piper --model your_model_name.onnx --output_file output.wav

Python API You can also use this model within a Python environment:

Python

import subprocess

def synthesize_albanian(text, model_path, output_path):
    command = f"echo '{text}' | piper --model {model_path} --output_file {output_path}"
    subprocess.run(command, shell=True)

synthesize_albanian("Si jeni sot?", "sq_AL-model.onnx", "test.wav")

Features Low Latency: Optimized for real-time synthesis even on low-end hardware (Raspberry Pi, etc.). Small Footprint: The ONNX format ensures the model is compact and portable. Natural Phonemes: Specifically tuned for Albanian phonology and characters like ë and ç.

Files sq_AL-medium.onnx: The core model weights.

sq_AL-medium.onnx.json: Configuration, phoneme map, and synthesis settings.

Disclaimer: This model is provided "as-is". Please ensure you credit the original rhasspy/piper-voices project if you redistribute this work.