Fidelity standard
How we verify
"100% true to the source texts" can't mean "we typed carefully." Every musical fact in Karunattu is one of three things: machine-verified against a formal rule, human-verified against a named source with the citation shown, or explicitly disclosed as an unverified placeholder. Nothing is silently presented as authoritative when it isn't.
72-melakarta table
Machine-verifiedAll 72 generated from the 17th-century mēḷakarta scheme and checked programmatically against the closed-form chakra/variant rules — zero deviations. Note what this verifies: internal consistency, not antiquity. The scheme is modern practice, four hundred years after the Saṅgīta Ratnakara; the arithmetic is exact and the scheme is not the treatise's.
Kaṭapayādi numeral cipher
Machine-verified71 of 72 melakarta names decode to their number under the documented consonant table and reading rules; Naṭabhairavī is a known historical anomaly, disclosed rather than hidden. See the Kaṭapayādi Explorer.
Sūḷādi tāḷa arithmetic
Machine-verifiedAll 35 jāti forms and 175 gati variants verified against closed-form aṅga math (Dhruva 3j+2, Rūpaka j+2, etc.) — zero failures. Modern practice, and labelled as such wherever the matrix is presented: the Saṅgīta Ratnakara does not describe the sūḷādi system. Its own tāḷa material is the mārga and deśi tāḷas, of which we carry none — see the coverage table above.
22-śruti / just-intonation cents
Machine-verifiedEvery default cent value matches its exact 5-limit ratio to within ±0.02¢. Per-rāga overrides use real alternative śrutis, each cited (SSP, Rajagopalan 1990).
Janya rāga scale degrees
Human-verified, citedEvery curated janya carries an audit record — a named source, correction notes, and a last-verified date — checked against Sambamoorthy's South Indian Music and cross-referenced with the app's own melakarta table via a permanent CI gate.
Kriti rāga/tāḷa/mātrā metadata
Human-verified, cited157 kritis' rāga and tāḷa labels are spot-verified against standard references; the notation layer runs a mātrā-count invariant gate with zero exceptions.
Piḷḷāri geethams (all 10)
Partially verified — disclosedRāga, tāḷa, composer, and the pallavi sahitya text for every geetham are now confirmed against 3+ independent published sources. The note-by-note melody is our own scale-conformant remapping of those real words — not a verified transcription of the authentic rhythm. Every geetham page discloses this explicitly.
Varnams (all 5)
Partially verified — disclosedRāga, tāḷa, composer, and Pallavi sahitya are confirmed per-varnam against published sources. The Anupallavi, Mukthāyi Svara, Caraṇam, and Ciṭṭa Svara sections remain unverified placeholders pending a full primary-source transcription — disclosed on every varnam page.
Gamaka Lab's 26 ornaments
Partially verified — disclosed17 are attested in Saṅgīta Ratnākara's 15 gamakas or the SSP's daśavidha 10 (labeled accordingly). The other 9 are contemporary pedagogical terms, marked "Contemporary" rather than attributed to a classical text.
What we carry, and what we don't
Ranking the rows by how much of each corpus we hold. The thin and empty ones are listed first, deliberately — depth in a few areas is not authority across all of them, and a list ordered to flatter itself would be marketing rather than provenance. Every count below is read from the live catalog, so adding a rāga updates this page.
Deśi tāḷas
0 of 120The treatise sets out 120 deśi tāḷas — its own fifth chapter's material. Neither this site nor the app carries any of them. What both carry is the 175-form sūḷādi matrix, which is modern. The platform is therefore deepest in a tāḷa system the Ratnākara did not describe, and empty on the one he did.
Mēḷakartas
72 of 72The 72-mēḷakarta scheme is modern practice — 17th-century, four hundred years after the Saṅgīta Ratnakara. Venkaṭamakhin formulated it; Govindacharya's naming is the order in use today.
Tāḷa forms (sūḷādi matrix)
175 of 1757 sapta × 5 jāti × 5 gati. This matrix is post-Ratnākara modern practice — the treatise does not describe it, and every surface presenting it carries that label.
Śrutis (named)
22 of 22All twenty-two are named. The names themselves are the treatise's; the cent values beside them are a modern just-intonation reconstruction. The treatise gives counts and names, not frequencies.
Gamakas
0The treatise names fifteen gamakas. This site teaches them inside lessons and reference profiles rather than as a browsable catalog, so it carries no ornament count of its own — the iOS app's Gamaka Lab is where the curated set lives. Nothing here should be read as an ornament inventory.
Synthesised kriti audio
0This site ships kṛti text and notation; it does not synthesise or stream audio. The iOS app does, and labels each melody as hand-transcribed or grammar-generated. Nothing here should be mistaken for a recording — no archival master is served through this site.
Rāgas
139The treatise describes 253 rāgas in its own classification. What the web carries is the 72-mēḷakarta system and its janyas — a later taxonomy on a different organising principle, so the two numbers count different things and no fraction is claimed here.
Totals come from the treatise where it gives one. Where it does not, no total is shown rather than an invented one — an invented denominator is the same kind of error as an invented citation. 1 of 7 corpora are carried only in part.
CI gates keeping this honest
- Every janya rāga's svaras must be a subset of its parent melakarta, unless declared bhāṣāṅga with named anya svaras.
- Every practice phrase's svaras must be a subset of its rāga's own scale.
- Every verified kriti's mātrā count must match its tāḷa cycle — zero exceptions.
- Sañcāra practice phrases are mechanically derived from already-verified kriti notation, not hand-authored separately — so provenance carries through automatically.
These run on every change, so a regression fails the build before it ships — not after a teacher notices.
Found something that doesn't match a source you trust? Tell us — a documented correction is worth more than a silent guess.