AVS 6.49.2

Source

Vedavani (Hugging Face)

Recording covers

one verse · AVS 6.49.2

Matched to this passage

Confirmed. The text the source says this recording recites matches this corpus's text for this passage.

The line behind the control

An accent trace: one turn for each of the 5 accent marks printed in this verse, a mark above the letter turning the line up and a mark below turning it down. It is derived from the text on this page and from nothing else. It is not measured pitch, not a reconstructed melody, and not a timing of the recording: the playhead moves across it because both run left to right, and no syllable is claimed to align with any moment of the audio.

Recitation tradition

Not stated by the source

Reciter

Not stated by the source

How it was matched

The dataset publishes no kanda.sukta.mantra key, so the key was established from the sample's own published transcript and the dataset's own ordering, and from nothing else. Four conditions had to hold together: the transcript is letter-identical to this corpus's Saunaka text for VG:AV:SAU:K06:S049:V002; the sample is filed under kanda 6, which is this mantra's kanda; its running index 246 falls between the indices of this mantra's nearest letter-identical neighbours in the same kanda, so the dataset's own ordering agrees with the text match; and it is the only sample in that window reciting this text. The boundary is the file boundary -- one standalone WAV containing exactly this mantra -- so no timestamp is asserted. No ASR was run: the transcript is the dataset's own published metadata.

Attribution

Vedavani: A Benchmark Corpus for ASR on Vedic Sanskrit Poetry (Kumar, Ray, Beerukuri, Kamoji, Jagadeeshan and Goyal, ACL 2025, arXiv:2506.00145). Apache License 2.0.

Licence

Apache-2.0

The dataset names no reciter and states no origin for the recording, so performer is null rather than guessed. The train split row for this file gives its duration as 12.937s, and 20 of 20 accepted rows were fetched and confirmed to be real RIFF/WAVE whose decoded duration matches that figure to within 0.02s. No row in this artifact was reviewed by listening. Admitted 2026-09-19 under OWNER_DECISION_AUDIO_TWO_TIER_PUBLICATION as SOURCE_MAPPED_UNREVIEWED. Media re-fetched that day and confirmed to resolve: content-type audio/wave, 414028 bytes. Staging evidence retained: patha_type=SAMHITA; text_similarity=1.0; quality_class=COMMUNITY_ARCHIVE; evidence_layer=DETERMINISTIC_DERIVED.