ukrainian-tts 6.0.2

Ukrainian TTS using ESPNET

---

title: "Ukrainian TTS" emoji: 🐌 colorFrom: blue colorTo: yellow sdk: gradio sdk_version : 5.7.1 python_version: 3.10.3 app_file: app.py pinned: false

Ukrainian TTS 📢🤖

Ukrainian TTS (text-to-speech) using ESPNET.

Link to online demo -> https://huggingface.co/spaces/robinhad/ukrainian-tts
Note: online demo saves user input to improve user experience; by using it, you consent to analyze this data.
Link to source code and models -> https://github.com/robinhad/ukrainian-tts
Telegram bot -> https://t.me/uk_tts_bot

Features ⚙️

• Completely offline
• Multiple voices
• Automatic stress with priority queue: acute -> user-defined > dictionary > model
• Control speech speed
• Python package works on Windows, Mac (x86/M1), Linux(x86/ARM)
• Inference on mobile devices (inference models through espnet_onnx without cleaners)

Support ❤️

If you like my work, please support ❤️ -> https://send.monobank.ua/jar/48iHq4xAXm
You're welcome to join UA Speech Recognition and Synthesis community: Telegram https://t.me/speech_recognition_uk

Examples 🤖

Oleksa (male):

https://github.com/robinhad/ukrainian-tts/assets/5759207/ace842ef-06d0-4b1f-ad49-5fda92999dbb

More voices 📢🤖

Tetiana (female):

https://github.com/robinhad/ukrainian-tts/assets/5759207/a6ecacf6-62ae-4fc5-b6d5-41e6cdd3d992

Dmytro (male):

https://github.com/robinhad/ukrainian-tts/assets/5759207/67d3dac9-6626-40ef-98e5-ec194096bbe0

Lada (female):

https://github.com/robinhad/ukrainian-tts/assets/5759207/fcf558b2-3ff9-4539-ad9e-8455b52223a4

Mykyta (male):

https://github.com/robinhad/ukrainian-tts/assets/5759207/033f5215-3f09-4021-ba19-1f55158445ca

How to use: 📢

Quickstart

Installation

Option 1: Using pip (Recommended)

pip install ukrainian-tts

Option 2: Using uv (Fast & Modern)

uv add ukrainian-tts

Option 3: Development Installation

# Using pip
pip install ukrainian-tts[dev]

# Using uv
uv add ukrainian-tts[dev]

Features Included

• ✅ Self-contained: No external git dependencies required
• ✅ All stress methods: Both dictionary and model-based stress
• ✅ Multiple voices: 5 different Ukrainian voices
• ✅ Cross-platform: Works on Windows, macOS, and Linux
• ✅ Fast installation: Optimized for modern Python package managers

Alternative Installation Methods

From Source (Development)

git clone https://github.com/robinhad/ukrainian-tts.git
cd ukrainian-tts
pip install -e .

Using uv for Development

git clone https://github.com/robinhad/ukrainian-tts.git
cd ukrainian-tts
uv pip install -e .

Code example:

from ukrainian_tts.tts import TTS, Voices, Stress
import IPython.display as ipd

tts = TTS(device="cpu") # can try gpu, mps

# Option 1: Save to file
with open("test.wav", mode="wb") as file:
    _, output_text = tts.tts("Привіт, як у тебе справи?", Voices.Dmytro.value, Stress.Dictionary.value, file)
print("Accented text:", output_text)

# Option 2: Get audio as variable (numpy array)
audio_array, sample_rate, accented_text = tts.tts_to_array("Привіт, як у тебе справи?", Voices.Dmytro.value, Stress.Dictionary.value)
print("Audio shape:", audio_array.shape)
print("Sample rate:", sample_rate)
print("Accented text:", accented_text)

# Option 3: Get audio as bytes
audio_bytes, accented_text = tts.tts_to_bytes("Привіт, як у тебе справи?", Voices.Dmytro.value, Stress.Dictionary.value)
print("Audio bytes length:", len(audio_bytes))
print("Accented text:", accented_text)

# Play audio in Jupyter
ipd.Audio(audio_array, rate=sample_rate)

Audio Output Methods 🎵

The Ukrainian TTS library provides three flexible ways to get audio output:

1. File Output (Traditional)

# Save to file
with open("output.wav", "wb") as file:
    _, accented_text = tts.tts("Привіт!", Voices.Dmytro.value, Stress.Dictionary.value, file)

2. Numpy Array (For Processing)

# Get raw audio data as numpy array
audio_array, sample_rate, accented_text = tts.tts_to_array("Привіт!", Voices.Dmytro.value, Stress.Dictionary.value)

# Perfect for:
# - Audio analysis and processing
# - Machine learning pipelines
# - Real-time audio manipulation
# - Jupyter notebook playback

3. Bytes Output (For APIs)

# Get WAV file as bytes
audio_bytes, accented_text = tts.tts_to_bytes("Привіт!", Voices.Dmytro.value, Stress.Dictionary.value)

# Perfect for:
# - Web APIs and HTTP responses
# - Database storage
# - Streaming applications
# - Microservices

4. Memory Buffer (BytesIO)

# Get file-like object in memory
output_buffer, accented_text = tts.tts("Привіт!", Voices.Dmytro.value, Stress.Dictionary.value)
audio_bytes = output_buffer.getvalue()  # Extract bytes when needed

# Perfect for:
# - When you need file-like interface
# - Temporary storage
# - Integration with other libraries

Use Case Examples:

🎯 Web API:

@app.route('/synthesize', methods=['POST'])
def synthesize():
    text = request.json['text']
    voice = request.json['voice']
    audio_bytes, _ = tts.tts_to_bytes(text, voice, Stress.Dictionary.value)
    return Response(audio_bytes, mimetype='audio/wav')

🔬 Audio Analysis:

audio_array, sample_rate, _ = tts.tts_to_array("Привіт!", Voices.Dmytro.value, Stress.Dictionary.value)
# Analyze audio features, apply filters, etc.

📱 Real-time Applications:

# Stream audio chunks
for chunk in audio_array.reshape(-1, 1024):  # Process in chunks
    # Send to audio output
    pass

See example notebook: tts_example.ipynb

macOS Installation 🍎

Simple Installation (Recommended)

The package is now self-contained and doesn't require system dependencies for basic usage:

# Using pip
pip install ukrainian-tts

# Using uv (faster)
uv add ukrainian-tts

Note: The package now includes all dependencies and works out-of-the-box! No system dependencies required for basic usage.

For Development/Advanced Usage

If you need to build the package from source or encounter issues, you can use our automated installation script:

git clone https://github.com/robinhad/ukrainian-tts.git
cd ukrainian-tts
./install.sh

This script handles:

• ✅ System dependencies (SentencePiece, CMake, pkg-config)
• ✅ Python virtual environment setup
• ✅ Package installation and testing

Troubleshooting

Flash Attention Warning:

Failed to import Flash Attention, using ESPnet default: No module named 'flash_attn'

This warning is normal on macOS and can be safely ignored. Flash Attention is designed for NVIDIA GPUs and not available on macOS.

System Dependencies (if needed):

brew install sentencepiece cmake pkg-config libsndfile
export PKG_CONFIG_PATH="/opt/homebrew/lib/pkgconfig:$PKG_CONFIG_PATH"

How to contribute: 🙌

Look into this list with current problems: https://github.com/robinhad/ukrainian-tts/issues/35

How to train: 🏋️

Link to guide: training/STEPS.md

Attribution 🤝

• Model training - Yurii Paniv @robinhad
• Open Source Ukrainian Text-to-Speech dataset - Yehor Smoliakov @egorsmkv
• Dmytro voice - Dmytro Chaplynskyi @dchaplinsky
• Silence cutting using HMM-GMM - Volodymyr Kyrylov @proger
• Autostress (with dictionary) using ukrainian-word-stress - Oleksiy Syvokon @asivokon
• Autostress (with model) using ukrainian-accentor - Bohdan Mykhailenko @NeonBohdan + Yehor Smoliakov @egorsmkv