screencastgen

Generate synchronized highlighting

The highlight pipeline adds word-level alignment to synthesized narration. PDF inputs use the original page images and word positions; other supported formats fall back to rendered text.

EPUB reader output

EPUB is the default output format:

screencastgen highlight book.pdf --format epub -o book-highlight.epub

MP4 output

screencastgen highlight book.pdf \
  --format mp4 \
  --resolution 1280x720 \
  --fps 24 \
  -o book-highlight.mp4

The local path requires the Qwen and WhisperX dependencies. To offload model work, add --backend remote --tts-server-url http://gpu-vm:8100.

Use --font-size to tune fallback text rendering. PDF page-image rendering keeps the source layout and matches aligned words to their original bounding boxes.

See Pipeline overview for processing stages and the Highlight Pipeline reference for implementation details.