Whisper Lecture Transcriber

A lightweight, fully local tool to transcribe university lectures or seminars (audio or video) into clean, editable text using OpenAI Whisper.

The process is fully automated:

• Sets up a Python virtual environment
• Installs all dependencies
• Transcribes any lecture file you provide
• Saves the transcript automatically in the transcripts/ folder

---

1. Setup Environment

Run this command once to prepare everything:

bash setup.sh

It will:

• Create a Python virtual environment (venv/)
• Install openai-whisper, torch, and torchaudio
• Verify that Python is properly available

---

2. Transcribe a Lecture

After setup, start the transcription process:

bash transcribe.sh

When prompted, type the name of your audio or video file (it must be in the same folder as the script):

Enter the name of the lecture file to transcribe (e.g., administrative_law_lecture.mp4): administrative_law_lecture.mp4

Whisper will process the file and generate a transcript in:

transcripts/administrative_law_lecture.txt

---

Project Structure

lecture-transcriber/
├── setup.sh           # Creates virtual environment and installs dependencies
├── transcribe.sh      # Runs Whisper transcription
├── requirements.txt   # Python dependencies
├── transcripts/       # Folder for generated transcripts
├── .gitignore
└── README.md

---

Supported File Formats

Whisper accepts all major audio and video formats, including:

.mp3, .mp4, .m4a, .wav, .ogg, .flac, .webm, .aac, .wma, .mkv

---

Model and Performance

• Default model: small (fast and accurate for English)
• To improve accuracy, edit transcribe.sh and replace:

  --model small

  with:

  --model medium

• Models available: tiny, base, small, medium, large
• Requires Python 3.9+

---

Troubleshooting

Q: “File not found”
➡ Make sure the file is in the same directory as transcribe.sh and that you typed the full name with the extension.

Q: “Virtual environment not found”
➡ Run bash setup.sh again to recreate it.

Q: Transcription too slow
➡ Try using a smaller model (tiny or base) for faster processing.

---

License

This project is released under the MIT License (© 2025 Baldari.dev).
It uses open-source components:

• OpenAI Whisper — MIT License
• PyTorch — BSD License

See the LICENSE file for full details.