Writing in Tamil, Malayalam,
and Hindi with Pinogrammer

How Pinogrammer detects and corrects Indic script text natively — on-device, in Tamil Unicode, Malayalam Unicode, Devanagari, Telugu, and Kannada — without any transliteration or cloud call.

Why Indian language support is different

Most browser grammar tools are built for English and Western European languages. When a Tamil, Malayalam, or Hindi speaker types in their native script on Gmail or WhatsApp Web, the tools either ignore the text entirely or produce nonsensical suggestions because they're treating Devanagari or Tamil Unicode as garbled input.

Pinogrammer was built with Indian scripts as first-class targets, not afterthoughts. The language detection pipeline treats Tamil Unicode (U+0B80–U+0BFF), Malayalam Unicode (U+0D00–U+0D7F), and Devanagari (U+0900–U+097F) as distinct script families, not "non-English text to skip". This means corrections work in native script — the suggestion card will show போகிறீர்கள்சென்றீர்கள், not a transliteration or an error about unrecognised characters.

How language detection works for Indian scripts

When you start typing in a text field, Pinogrammer samples the first significant cluster of text (after 650ms of inactivity) and passes it through two detection layers:

  1. Unicode block analysis: The majority Unicode block in the text identifies the script family. If the dominant block is Tamil Unicode, the language is flagged as Tamil. If Devanagari, it's flagged as Hindi (or optionally Marathi, based on vocabulary). This runs locally in about 2ms — no model call needed.
  2. Chrome Language Detector API: For Latin-script languages (English, French, Spanish etc.) and for confirmation of Indic languages in mixed-language text, Pinogrammer calls Chrome's built-in LanguageDetector API. This is also on-device and returns a confidence score.

The script detected is shown in the suggestion card header as a language badge (e.g., "Tamil detected", "Malayalam detected"). You can see which path was used — Gemini Nano or Translate + Nano — in the card's meta line.

Tamil and Malayalam are the most thoroughly tested Indic scripts in Pinogrammer because they have strong grammatical structure that maps well to Gemini Nano's training data. Hindi, Telugu, and Kannada follow closely. Scripts like Gujarati, Punjabi, and Bengali use the Chrome Translator bridge (translate to English → correct → translate back) but this remains entirely on-device.

Correction examples by language

Tamil
தமிழ்
நான் நேற்று போகிறேன் சென்றேன்
Verb tense: present → past
அவர்கள் வருகிறார்கள் வந்தார்கள் நேற்று
Tense agreement with temporal marker
Malayalam
മലയാളം
ഞാൻ പോകുന്നു പോയി ഇന്നലെ
Past tense with temporal adverb
അവർ ആണ് ആയിരുന്നു കൂടെ
Copula agreement in past context
हि
Hindi
हिन्दी
मैं कल जाता हूँ गया था
Past tense with temporal marker
वह बहुत अच्छा बहुत अच्छी लड़की है
Gender agreement: masculine → feminine
తె
Telugu
తెలుగు
నేను నిన్న వెళ్తున్నాను వెళ్ళాను
Present continuous → simple past
ఆమె మంచి మంచిగా పాడింది
Adjective → adverb before verb

Setting up an Indian language keyboard

Pinogrammer corrects Unicode text — it needs your keyboard to output Unicode codepoints in the native script, not transliterated Latin text. Here are the recommended setups per platform:

LanguagemacOSWindowsAndroid/Chrome
Tamil Tamil keyboard (System Preferences → Keyboard → Input Sources) Microsoft Tamil keyboard (Language settings) Google Indic Keyboard → Tamil
Malayalam Malayalam keyboard (Input Sources) Microsoft Malayalam keyboard Google Indic Keyboard → Malayalam
Hindi Hindi — Devanagari QWERTY or Transliteration Microsoft Hindi (Devanagari) Gboard → Hindi
Telugu Telugu keyboard (Input Sources) Microsoft Telugu Google Indic Keyboard → Telugu
Kannada Kannada keyboard (Input Sources) Microsoft Kannada Google Indic Keyboard → Kannada

Google Input Tools (Chrome extension) is another excellent option for all Indic scripts. It supports phonetic input (type "naan" → converts to நான்) and outputs native Unicode, which is exactly what Pinogrammer reads.

Writing in mixed English and Indian languages

A common pattern for Indian professionals is Tanglish (Tamil + English), Manglish (Malayalam + English), or Hinglish (Hindi + English) — sentences that mix native script words with English words in the same message.

Pinogrammer handles this by segment: if a sentence has Tamil words and English words, the Tamil portion is corrected using Tamil grammar rules and the English portion using English rules. The boundary detection uses script switching as the segment boundary.

Example Gmail compose: "Dear sir, நான் tomorrow meeting க்கு வர முடியாது என்று சொல்கிறேன்." — The English "Dear sir" and "tomorrow meeting" are checked for English grammar. The Tamil segments are corrected independently with Tamil grammar rules.

💡
Transliteration note: If you write Tamil phonetically in Latin characters (e.g., "naan tomorrow meeting ku vara mudiyaadu") Pinogrammer will treat this as (potentially misspelled) English and may make incorrect suggestions. Switch to native script input for best results with Indic language correction.

Offline support for Indic scripts

By default, Tamil and Malayalam corrections use Gemini Nano — Chrome's on-device model. This works fully offline once the model is downloaded. If you frequently write in scripts that use the Translate + Nano path (Bengali, Punjabi, Gujarati), you can opt into the M2M100 model download in Pinogrammer Settings → Language Models. This is a one-time 170MB download that enables fully offline multilingual correction for all 46 supported languages.

To enable: open the Pinogrammer settings popup → Language tab → "Download offline model" button. Chrome will prompt you once to confirm the download. After that, correction works identically whether or not you have an internet connection.

Tips for Indic language writers

Complete guide → Domain profiles → Pinogrammer overview →