Text Shaping

Text shaping is the process by which software converts a sequence of Unicode text characters into positioned glyphs ready for rendering. For simple scripts like English, shaping is trivial — each letter maps to one glyph placed left-to-right. For complex scripts (Devanagari, Tamil, Malayalam, Arabic Nastaliq, Thai), shaping is far more involved: the shaping engine must form conjuncts (combining consonants into ligatures), reorder visual elements (Indic matra ि that's typed after but displayed before), substitute contextual letter forms (Arabic letters have 2-4 shapes depending on position), position diacritics correctly above or below base letters, and apply kerning. HarfBuzz is the open-source industry-standard text shaping engine — used by Android, Firefox, Chrome, Apple platforms, and most Linux distributions. Software that doesn't integrate HarfBuzz (or an equivalent shaper) renders complex scripts incorrectly: visible halants in Indic text, isolated letter forms in Arabic, mis-positioned matras. Quality text shaping is invisible when it works and obvious when it fails.

How it works

A shaping engine reads input as Unicode code points, looks up the font's OpenType shaping tables (GSUB — glyph substitution, GPOS — glyph positioning), and produces a sequence of glyph IDs with (x, y) offsets. For Indic scripts, this process typically runs: detect the script (Devanagari, Tamil, etc.), identify syllable clusters (akshara — consonant + optional matra), apply syllable-level rules (handle halant-based conjuncts, handle above/below/post/pre matras), position glyphs. For Arabic/Nastaliq: identify word-initial/medial/final letter positions, look up contextual alternates, apply ligatures (some 2-to-1, some 3-to-1 or larger), kern neighbouring letters. Modern fonts ship with extensive OpenType tables — e.g., Noto Sans Devanagari has thousands of GSUB/GPOS rules covering common conjuncts, rare conjuncts, and edge cases. Software without full OpenType shaping falls back to rendering each glyph as-stored — producing technically-readable but visually-wrong output for complex scripts.

Examples

Related terms

Learn more

Frequently Asked Questions