Vāgdhenu is a text-to-speech system designed specifically for converting Sanskrit metrical verses (ślokas) into chant audio, with a focus on accurate metrical (vṛtta) awareness. Developed and maintained by Prof. Prathosh at the Indian Institute of Science, Bengaluru, the tool addresses the challenge of faithfully rendering Sanskrit ślokas in their traditional chanted forms, preserving both linguistic and metrical features.
The system automatically detects the meter of a Sanskrit verse pasted in any Indian script and generates a chanted audio version. Its architecture includes a flow-matching TTS backbone retrained on a purpose-recorded, single-speaker Sanskrit chant corpus of approximately five hours, with further voice-steering retraining. The neural vocoder is fine-tuned specifically for the chant register. Vāgdhenu features a script-aware frontend that routes Sanskrit text through Kannada orthography, which is noted to avoid certain issues present in Devanagari script processing, such as schwa-deletion. The tool handles visarga sandhi with the correct allophones (jihvāmūlīya and upadhmānīya), maintains aspiration contrasts, and preserves distinctions among the three sibilants and the full retroflex series. It also supports homorganic anusvāra and vocalic ṝ, and includes a vṛtta-aware mechanism that selects a matched reference under the half-reference rule for accurate metrical chanting.
Vāgdhenu has been used to produce large-scale chant corpora, including 32 chapters of Mahābhārata Tātparya Nirṇaya (over 5,000 verses) and the complete Śrīmad Bhāgavatam (about 18,000 verses across 12 books). The system is accessible through a web-based demo where users can input Sanskrit verses and listen to the generated chants. Additionally, its recitations power the Bhāgavata-VāNi app, which offers synced audio recitation and karaoke-style highlighting in 10 Indian scripts, with search and offline functionality.
0. The system builds on AI4Bharat IndicF5 and NVIDIA BigVGAN-v2 base models, which are MIT-licensed. Vāgdhenu stands as a specialized tool within the class of Sanskrit chant TTS systems, offering open access to both its software and datasets.
Vāgdhenu sits in PulseGate's Voice, TTS & speech category. It focuses on automating the chanting of Sanskrit ślokas with accurate meter and pronunciation using TTS. It is built as an open-source project for sanskrit scholars, linguists, and developers. Vāgdhenu is open source under the Apache-2.0 license. It runs on the web.
Prof. Prathosh A P, Indian Institute of Science builds and maintains Vāgdhenu, and the product first shipped in 2026. Development happens publicly on GitHub with 137 stars and 24 commits in the last 90 days. Key capabilities include Sanskrit TTS, meter detection, and chant synthesis.
Latest indexed changes and source events
Vāgdhenu: Open-source meter-aware text-to-speech for Sanskrit verified by the PulseGate indexer
Other apps tracked under the same category.
