In the Japanese writing system, hentaigana (変体仮名, 変体がな; Japanese pronunciation: [hentaiɡana] or [hentaꜜiɡana], lit. "variant kana") are variant forms of hiragana.

In contrast to modern Japanese, hiragana historically had several distinct forms representing a single sound. For example, while the hiragana reading "ha" has only one form in modern Japanese (は), until the Meiji era (1868–1912) it was written in various forms, including , , and . The shift to using only one character for each sound occurred as part of the 1900 script reform, which also included other changes to the written language to standardize spelling (and was part of a larger project to westernize the country).

As a result of this state-mandated standardization of hiragana, variant kana have fallen into disuse in modern Japan, save for limited situations such as signboards, calligraphy, place names, and personal names. Today, those hiragana glyphs not used in school education since 1900 as a result of the script reform are called hentaigana.

Hiragana, the main Japanese syllabic writing system, derived from a cursive form of man'yōgana, a system where Chinese ideograms (kanji) were used to write sounds without regard to their meaning. Originally, the same syllable (more precisely, mora) could be represented by several more-or-less interchangeable kanji, or different cursive styles of the same kanji. However, the 1900 script reform determined that only one specific character be used for each mora, with the rest being called hentaigana ("variant characters").

The 1900 standard included the hiragana ゐ, ゑ, and を, which historically stood for the phonetically distinct moras /wi/, /we/, and /wo/ but are currently pronounced as /i/, /e/, and /o/, identically to い, え, and お. These characters were deprecated by the 1946 spelling reform, except for the character を, which is still used in the Japanese writing system for the direct object particle /-o/.

Hentaigana are still used occasionally today in some contexts, such as store signs and logos, to achieve the "old-fashioned" or "traditional" look.

Katakana also has variant forms, such as (ネ) and (ヰ). However, katakana's variant forms are fewer than hiragana's. Katakana's choices of man'yōgana segments had stabilized early on and established – with few exceptions – an unambiguous phonemic orthography (one symbol per sound) long before the 1900 script regularization.

Before the proposal which led to the inclusion of hentaigana in Unicode 10.0, they were already standardized into a list by Mojikiban, part of the Japanese Information-technology Promotion Agency (IPA).