🎤 Audio Transcription & Speaker Diarization

Transcribe Large audio files with automatic speaker diarization and a web interface.

Uses Nvdia Parakeet v2 and Pyannote models for transcription and diarization.

---

🚀 Features

• Automatic Speech Transcription (English)
• Speaker Diarization: Distinguish between multiple speakers
• Chunked Processing: Handles long audio files efficiently
• Speaker Recognisation: Correctly recognizes these speakers across all the chunks.
• Modern Web UI: Powered by Gradio
• Downloadable Results: Get both formatted text and grouped-by-speaker files
• CPU Only Infernce: No need for GPU Uses PyTorch and ONNX for fast inference GPU is optional.

---

📸 Demo

---

🛠️ Setup Instructions

1. Clone the Repository

git clone https://github.com/deepanshu-yadav/big_audio_file_transcription.git
cd big_audio_file_transcription

---

2. Download Pretrained Models

Diarization Models

If you choose pyannote models for diarization you need to download the following files. Otherwise if you choose resemblyser then no need.

• Segmentation Model
  Download from pyannote/segmentation-3.0
  Then run:

  cp pytorch_model.bin model_components/pyannote/
  mv model_components/pyannote/pytorch_model.bin model_components/pyannote/segmentation-3.0.bin

• Embedding Model
  Download from pyannote/wespeaker-voxceleb-resnet34-LM
  Then run:

  cp pytorch_model.bin model_components/pyannote/
  mv model_components/pyannote/pytorch_model.bin model_components/pyannote/wespeaker-voxceleb-resnet34-LM.bin

Transcription Model

Download and extract:

wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2
tar xvf sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2
rm sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2

Copy all .onnx and tokens.txt files into the model_components folder.

---

3. Install Dependencies

Python Packages

pip install -r requirements.txt

FFmpeg (Required for audio processing)

• Ubuntu/Debian:
  sudo apt update && sudo apt install ffmpeg
• Arch Linux:
  sudo pacman -S ffmpeg
• macOS (Homebrew):
  brew install ffmpeg
• Windows (Winget):
  winget install Gyan.FFmpeg
• Windows (Chocolatey):
  choco install ffmpeg
• Windows (Scoop):
  scoop install ffmpeg

Or download from the official FFmpeg website.

---

💻 Usage

Start the Web App

python app.py

Visit http://127.0.0.1:7860 in your browser.

---

How to Use

1. Upload your audio file (supports .wav, .mp3, etc.).
2. Set the number of speakers and chunk duration.
3. Click "Transcribe".
4. View the transcription with speaker labels.
5. Download the full results as a text file.

---

📂 Project Structure

repo/
│
├── app.py                  # Gradio web interface
├── transcript_big_file.py  # Main diarization & transcription logic
├── transcript.py           # Feature extraction & ONNX model handling
├── utils.py                # Utility functions (audio conversion, file writing)
├── requirements.txt        # Python dependencies
├── model_components/       # Place all downloaded models here
│   └── pyannote/
│       ├── config_diarize.yaml
└── README.md

---

⚡ Future Improvements

• Support for languages other than English
• Improved handling of overlapping speakers

---

🤝 Contributing

Pull requests and issues are welcome! Please open an issue for bugs or feature requests.

---

📄 License

This project is licensed under the MIT License.

---

Enjoy fast, accurate, and speaker-aware audio transcription!