summarize_transcript

📝 Summarize Transcript

This Python tool generates a concise summary from a text transcript file. It supports transcripts in plain text (.txt) format or JSON format (specifically, the output from OpenAI Whisper).

🔧 Features

• Summarizes long texts into a shorter, coherent summary.
• Uses pre-trained summarization models from Hugging Face Transformers (e.g., BART, T5).
• Supports input from .txt files or Whisper-generated .json files.
• Allows customization of summary length (min/max tokens).
• Option to specify which summarization model to use.

🏁 Quickstart

# Install dependencies
pip install -r requirements.txt

# Summarize a plain text transcript
python summarize_script.py --input_file examples/sample_transcript.txt --output_file summary_plain.txt

# Summarize a Whisper JSON transcript
python summarize_script.py --input_file examples/sample_transcript.json --output_file summary_whisper.txt --model_name t5-small

# Specify summary length constraints
python summarize_script.py --input_file examples/sample_transcript.txt --output_file custom_summary.txt --min_length 30 --max_length 150

⚙️ How it Works

1. Load Transcript: The script loads the input transcript. If it's a JSON file from Whisper, it extracts the full transcribed text. If it's a .txt file, it reads the content directly.
2. Load Model: A pre-trained summarization model (e.g., facebook/bart-large-cnn by default, or another specified by the user) is loaded from the Hugging Face Transformers library.
3. Summarization: The transcript text is fed into the model, which generates a summary.
4. Output: The generated summary is saved to the specified output text file.

📂 Input Files

• .txt: A plain text file containing the transcript.
• .json: A JSON file in the format output by OpenAI Whisper (must contain a "text" field with the full transcript or "segments" array).

📄 Output File

A plain text file containing the generated summary.