vntts 0.1.5

Vietnamese Text-to-Speech and Speech-to-Text library with embedded models

VNTTS

Dự án tích hợp Vietnamese Text-to-Speech (TTS) và Speech-to-Text (STT):

• TTS: nghitts / Piper-compatible ONNX (giọng Banmai)
• STT: sherpa-onnx-zipformer-vi-2025-04-20

Cấu trúc dự án

VNTTS/
  src/vntts/
    __init__.py
    tts.py
    stt.py
    model_downloader.py   # Downloads models from GitHub Releases
    model_parts.py        # Helpers for split/merge model files
  scripts/
    split_model.py
    merge_model.py
  tests/
  README.md

Cài đặt

python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Model Loading

Models are automatically downloaded from GitHub Releases on first use. When TTS() or STT() is instantiated without an explicit model_dir, the library:

1. Checks if the models directory exists (default: ./models)
2. If not, downloads models.zip from the configured GitHub Release URL
3. Extracts the zip and removes the archive
4. Subsequent runs skip the download (idempotent)

Environment Variables

Variable	Description	Default
VNTTS_MODELS_DIR	Root directory for models	models
VNTTS_MODELS_URL	URL for models.zip	(GitHub Release)
VNTTS_MODEL_DOWNLOAD_TIMEOUT_SECONDS	Network timeout	120
VNTTS_MODEL_DOWNLOAD_RETRIES	Retry attempts	3
VNTTS_MODEL_DOWNLOAD_LOG	Set 0 to disable logs	1

Explicit Model Directory

You can also provide your own model directory:

from vntts import TTS, STT

tts = TTS(model_dir="/path/to/tts/models")
stt = STT(model_dir="/path/to/stt/models")

Testing

pytest

Building

python -m build