TranslateAI

TranslateAI is a powerful real-time speech translation desktop application built using PyQt and Hugging Face models. It enables users to convert spoken words into text and translate them into different languages.

MacOS	Windows

TranslateAI.mp4

This app allows you to:

• Select your preferred input device, whether it's a microphone or system audio, ensuring flexibility in capturing speech.
• Speak or even play pre-recorded audio files, and the app will process and transcribe the speech into text in real time.
• Enjoy an automatic translation trigger, which activates after about 1 second of silence, making the experience smooth and natural.
• Translate between multiple languages, including Turkish 🇹🇷, English 🇬🇧, Spanish 🇪🇸, and French 🇫🇷, with the possibility of adding more in the future.

Designed with ease of use and efficiency in mind, TranslateAI brings a fluid translation experience for various use cases, whether you're learning a new language, working in multilingual environments, or simply looking for an intuitive speech-to-text tool.

📌 Contents

• 🚀 Features
• 📦 Installation
  • MacOS
  • Windows
  • Linux
• 🛠️ Usage
• 🧠 How It Works
  • Speech Detection
  • Translation Model
• ⚙️ Configuration
• 🤝 Contributing
• 📝 License

---

🚀 Features

• 🎙️ Real-time Speech & Audio Translation
• 🔊 Visual Indicator for Sound Intensity
• 🌐 Multi-language Support: Turkish 🇹🇷, English 🇺🇸, Spanish 🇪🇸, French 🇫🇷 (More to come!)
• 🤖 Hugging Face Model Integration (Helsinki-NLP)
• ⚡ Automatic Model Download Based on Selected Language
• 🖥️ User-Friendly PyQt GUI

---

📦 Installation

1. Clone the Repository

git clone https://github.com/furkankarakuz/TranslateAI.git
cd TranslateAI

2. Set Up a Python Environment (Recommended)

It’s often best to use a virtual environment:

python -m venv venv
source venv/bin/activate  # On Mac/Linux
venv\Scripts\activate     # On Windows

3. Install Dependencies

Install required Python packages:

pip install -r requirements.txt

Note: If requirements.txt doesn’t exist or you prefer manual installation, make sure to install:

• PyQt5 or PyQt6 (depending on your code)
• pyaudio (or sounddevice / pydub if you adapt the code)
• torch, transformers, datasets (for Hugging Face)

MacOS

On macOS, you may need to install portaudio via Homebrew before installing pyaudio:

brew install portaudio

Then:

pip install pyaudio

Windows

On Windows, you can directly install pyaudio from PyPI:

pip install pyaudio

If you encounter issues, you might need to install the appropriate Microsoft Visual C++ Build Tools.

Linux

On Linux (Ubuntu/Debian-based), you may need:

sudo apt-get update
sudo apt-get install portaudio19-dev python3-pyaudio

Then install Python packages as usual:

pip install -r requirements.txt

---

🛠️ Usage

1. Run the Application

   python app.py

2. Select Your Audio Device

   • In the top section of the UI, choose your microphone or system audio input (e.g., “MacBook Air Microphone …”).

3. Choose Language Pair

   • Select the source and target language. For example, EN-TR means you will speak English and get Turkish translations.

4. Start Recording

   • Click Start Record to begin capturing audio.
   • Watch the volume indicator to see if audio is being detected.
   • After you stop speaking for about 1 second, the application will automatically trigger the translation.

5. View Translations

   • The translated text will appear in the console.

6. Stop Recording

   • Click Stop Record when finished.

---

🧠 How It Works

Speech Detection

• PyAudio captures your microphone input (or chosen device).
• A volume level meter is displayed to help you monitor input levels.
• Once silence (below a certain threshold) is detected for ~1 second, the app processes the captured audio chunk and sends it for transcription & translation.

Translation Model

• The application uses Hugging Face Transformers to download and run the appropriate translation model for the chosen language pair.
• Model caching: Once a model is downloaded, it should be reused for subsequent translations to save time.
• The translations are displayed in real-time, providing an instant feedback loop.

---

⚙️ Configuration

• Silence Threshold & Delay: Currently set to about 1 second. You can modify this value in the code if you want quicker or slower triggers.
• Language Support: To add a new language pair, you need to:
  • Find a Hugging Face translation model that supports that pair.
  • Update the UI to include the new language option.
  • Adjust the code that downloads/loads the model.

---

🤝 Contributing

Contributions are welcome! Feel free to:

• Fork this repository.
• Create a new branch.
• Commit your changes.
• Open a Pull Request describing the improvements or bug fixes you’ve made.

---

📝 License

This project is licensed under the Apache License 2.0