Complete video processing workflow, executed in three steps:
Step 1 + 2 - Screenshot and automatic OCR
source .venv/bin/activate && python pipeline.py process --save-dir output/captures --interval 500
Inform the user:
- Click "Select Area" to select the video area on the screen
- Click "Start Capture" to begin timed screenshots
- Play the video
- Screenshots will automatically end after stopping (automatically stops when no new screenshots exceed interval*2), then OCR will run automatically
Wait for the command to finish executing.
Step 3 - Generate article
Read output/extracted_text.txt, organize the scattered OCR text into a coherent Markdown article:
- Remove OCR noise (garbled text, UI text not part of the video content, etc.)
- Remove duplicate and meaningless content
- Organize paragraphs logically
- Language is fluent, preserving the original meaning
Write the final article to the output/article.md file, and inform the user of the file path.