Tulā Kāverī Māhātmya
Two reciters, recorded years apart, chant 141 of the same verses in the same order. Those verses appear in none of the 488,149 verses we index, and no digital exemplar of the text has been found. So this page does not put a printed edition against a recording, the way the atlases do. It puts one recording against another.
Every atlas on this site is text-first: a printed edition is the spine, the recording is hung on it, and the point worth making is that the two were authenticated separately. Here there is no spine. We could have built one out of the transcript and seated the audio against it — and every timestamp on the resulting page would have been true by construction, because the text side and the audio side would be the same measurement wearing two hats. That page would look exactly like the honest ones and would mean nothing.
What is left is better than it sounds. Two independent performances, transcribed by separate passes and aligned by separate passes, agreeing verse for verse, is evidence about a text — produced without the text. Each row below is one such agreement: the words as each witness decoded them, and the second each of them begins.
Nobody assigned these chapter numbers. The reciter reads the colophon out at the close of each adhyāya, and they run strictly ascending across the series — so the numbering below is quoted, not counted. The named division, , exists in neither circulating Agni Purāṇa: this is the ordinary signature of a regional māhātmya carrying a purāṇic colophon that no printed purāṇa matches.
Ordered as the continuous series walks the text. Press one voice, then the other: the same words, in two mouths, at two tempos. Where the two decodes differ, both are printed — the divergence brackets the reading rather than hiding behind one of them, and neither is an edition.