sprachText/README.md

# 🎙️ Voice-to-Text App with Whisper

This is a simple, user-friendly application that records your voice and converts it into text using OpenAI's Whisper model. Built with 💻 `Python`, 🎵 `sounddevice`, and 🤗 `Gradio`, this app is designed for local use, requiring internet access **only during the initial setup**.

---

## 🌟 Features

- 🎤 **Record Audio**: Click a button to start and stop recording your voice.
- ✂️ **Automatic Splitting**: Handles long audio files by splitting them into smaller chunks for transcription.
- 📝 **Speech-to-Text**: Transcribes your voice into text using the Whisper model.
- 🔒 **Offline Capability**: After setup, the app works entirely offline.

---

## 🚀 Getting Started

### Prerequisites
1. **Python 3.8+**
2. Install the required Python libraries:
   ```bash
   pip install torch transformers sounddevice pydub gradio
   pip install --upgrade transformers datasets[audio] accelerate  
   ```
3. **FFmpeg** (for audio processing):
   - Download and install FFmpeg from [FFmpeg Official Site](https://ffmpeg.org/download.html) or via [chocolately](https://chocolatey.org/).

4. **Possible Bugs:** If there are any problems with the GPU use:
    ```bash
    pip uninstall torch torchvision torchaudio  
    ```
    Search for the right torch version for you GPU and intall torch
    ```bash
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124 
    ```
---

### 📦 Installation

1. Clone this repository:
   ```bash
   git clone https://github.com/your-username/voice-to-text-app.git
   cd voice-to-text-app
   ```

2. Run the app:
   ```bash
   python app.py
   ```

3. Open the provided link in your browser to access the web app.

---

## 🛠️ How It Works

1. **Recording**: Click the **Start Recording** button to record your voice. Click **Stop Recording** when you're done.
2. **Transcription**: Click the **Transcribe** button to convert your audio to text.
3. **Automatic Handling**: The app automatically splits audio longer than 30 seconds and transcribes it in chunks.

---

## 🌐 Internet Usage

- **Setup**: The app requires an internet connection **only during the first run** to download the Whisper model and dependencies.
- **Offline Mode**: Once the model is downloaded, the app works entirely offline, ensuring privacy and local processing. (As long as the cache is not deleted. Otherwise the model will be downloaded again.)

---

## 🎉 Example Use Case

1. Record a 20-second audio note: "Take out the trash at 6 PM."
2. Stop recording.
3. Transcribe the audio to see: `"Take out the trash at 6 PM."`

---

## 🧰 Built With

- 🤗 [Transformers](https://huggingface.co/docs/transformers): Whisper model for speech-to-text.
- 🎵 [SoundDevice](https://python-sounddevice.readthedocs.io/): Audio recording.
- ✂️ [Pydub](https://github.com/jiaaro/pydub): Audio splitting for long files.
- 🌐 [Gradio](https://gradio.app/): Interactive web interface.

---

## 🤝 Contributing

Contributions are welcome! Feel free to submit issues or pull requests to improve this project.

---

## 🛡️ License

This project is licensed under the MIT License.

---

## 🗂️ File Structure

```
.
├── app.py             # Main application script
├── requirements.txt   # List of dependencies
└── README.md          # This file
```

---

Have fun 🤗
---