Which schwas do you actually pronounce?
Devanagari packs a vowel into every consonant, but spoken Hindi quietly drops many of them. This is why राम is read “Ram” and not “Rama,” and it trips up learners constantly because the dropped vowel is still written. This tool reads your Devanagari text and marks each inherent schwa as pronounced or deleted.
How it works
The tool first tokenizes each word into consonant units, noting for each whether it carries an inherent schwa, an explicit vowel sign, or a virama that cancels the schwa. Conjunct clusters joined by a virama are kept together. It then applies the two standard deletion rules:
- The final inherent schwa of a word is deleted.
- An internal inherent schwa is deleted when it stands in the context VC_CV — preceded by a vowel-bearing unit and followed by a consonant that has its own vowel.
The rule is applied right to left, the order in which Ohala’s analysis predicts Hindi syncope. A green underline marks a surviving schwa, a red underline a deleted one.
Notes and example
Take a word like भारत: the first syllable भा has an explicit long-a sign, र carries an inherent schwa, and the final त loses its schwa by the final-deletion rule, giving “Bharat.” Remember the prediction follows the regular rules only — loanwords, compounds, and names can break the pattern, so use the colors as a strong guide and trust your ear for exceptions. For related invisible-vowel rules in other scripts, see the Persian ezafe detector.