screencastgen

Highlight Renderer

PIL-based word-highlighted text frame renderer (plain-background fallback).

Source: screencastgen/highlight_renderer.py

Overview

Renders individual video frames showing text with the currently-spoken word highlighted on a plain dark background. Uses PIL/Pillow for drawing with system fonts and handles word wrapping, scrolling, and per-word color highlighting.

This is the fallback renderer used when the input is not a PDF or PyMuPDF is not installed. For PDF inputs, Page Renderer is preferred — it highlights words on the actual PDF page images.

Class: `HighlightRenderer`

Constructor

HighlightRenderer(width=1280, height=720, font_size=32)

Methods

Method	Description
`layout_words(words) -> List[dict]`	Computes pixel positions for each word. Accepts `List[str]` or `List[WordTiming]`.
`render_frame(layout, active_index, scroll_offset) -> Image`	Renders a single frame with the active word highlighted
`compute_scroll_offset(layout, active_index) -> int`	Calculates vertical scroll position to keep active word visible
`get_active_word_index(words, time) -> Optional[int]`	Finds which word is active at a given timestamp

Rendering Details

Font: System sans-serif (DejaVu Sans, Liberation Sans, etc.)
Background: Dark gray (30, 30, 30)
Text: White (255, 255, 255)
Highlight: Yellow background (255, 255, 0) on active word
Scrolling: Auto-scrolls to keep the active word in the viewport
Word wrapping: Breaks lines at word boundaries based on frame width

Dependencies

Highlight Renderer
├── Pillow (PIL)     (deferred import)
├── Types         (WordTiming for timing data)
└──▶ consumed by Highlight Pipeline
     └──▶ consumed by Video Composer

Data Flow

List[WordTiming]
    │
    ▼  layout_words()
List[dict] positions (x, y per word)
    │
    ▼  render_frame() (called per video frame)
PIL Image
    │
    ▼  passed to Video Composer

layout_words() also accepts List[str], but the current video pipeline passes AlignedChunk.words directly so both renderers share the same interface.