Vāgdhenu TTS

Chant Sanskrit verses in their traditional meter, any script.

Example output
0:00 / 0:00
Try an example3

Vāgdhenu TTS — Sanskrit Text-to-Audio

What is Vāgdhenu TTS?

Vāgdhenu TTS turns a single classical Sanskrit verse into a chanted recitation (pārāyaṇa): it sings the śloka in the rhythm of its meter rather than reading it out flatly. Vāgdhenu is the open model by Prof. Prathosh A P of the Indian Institute of Science, Bengaluru, served here as he released it. Paste a verse in any major Indian script and the model reads its meter, matches a traditional melodic contour and returns a metered chant in a single, purpose-recorded voice.

Key Features

  • Meter-aware chanting — auto-detects the meter (chandas) from a complete verse, or you select one of 15 named meters plus gadya (prose).
  • Ten Indian scripts in — Devanagari, Kannada, Telugu, Malayalam, Bengali, Gujarati, Gurmukhi, Odia, Tamil and Grantha. Romanised text is not accepted.
  • Faithful classical recitation — dense conjuncts, including retroflex aspirates, render correctly.
  • 24kHz mono audio as WAV or MP3, with an optional seed for reproducible takes.
  • Checkable output — the response reports the meter used and the syllable count.

Best Use Cases

Vāgdhenu TTS fits stotra and śloka apps, devotional and educational audio, chant-practice tools, karaoke-style verse highlighting and bulk recitation of long texts. The same system rendered two large corpora — the Mahābhārata Tātparya Nirṇaya (5,183 verses) and the complete Śrīmad Bhāgavatam (around 18,000 verses) — so it holds up across long, varied material. Because it takes whatever Indian script you already have, it serves multi-script audiences without a separate transliteration step.

Prompt Tips and Output Quality

Send one complete verse with correct line breaks — one half-verse per line, or halves separated by a daṇḍa (।) — and leave the meter on auto so the detector can read it from the full verse. For a partial verse, or when auto-detection guesses wrong, set the meter yourself. Paste the verse as written, keeping anusvāra (ṃ) and visarga (ḥ), and keep to classical (laukika) verse: it does not render Vedic accents. Fix a seed to reproduce a take exactly.