Skip to main content
Dataset catalog

Rights-cleared sovereign datasets

Proprietary, consented, rights-clean datasets for the world's under-served languages — low-resource and code-mixed languages across the Global South, licensed repeatably, with documented consent, provenance, measured quality (IAA ≥ 0.85), and India data residency.
Datasets live
41
Language footprint
400
Annotation IAA
≥ 0.85
Residency
India

Every dataset is rights-cleared with documented consent and a downloadable dataset card. Pricing is shown per licence tier on each catalog card below.

Showing 3 of 41 datasets

  • General
    General Audio / Speech 🇮🇳 Sovereign

    Low-Resource Language ASR & TTS Corpus

    Paired speech-and-text for low-resource languages, built for both ASR and TTS — read prompts plus spontaneous speech, with studio-grade single-speaker sets for voice building.

    Languages:
    Santali, Bodo, Dogri, Maithili + Global-South LRLs
    Size:
    410 hrs · 1,300 speakers · 12 languages
    IAA:
    0.87

    Licence from

    Contact us

    Rights-cleared
    View data card See pricing
  • Safety / Eval
    Safety / Eval Text 🇮🇳 Sovereign

    South Asian Instruction-Following Eval (Rare Languages)

    An instruction-following and reasoning evaluation suite for four under-served South Asian languages, with human reference answers and rubric-based scoring.

    Languages:
    Maithili, Bhojpuri, Santali, Sindhi
    Size:
    9,600 instruction–response pairs
    IAA:
    0.84

    Licence from

    Contact us

    Rights-cleared
    View data card See pricing
  • General
    General Audio / Speech 🇮🇳 Sovereign

    Rare-Language Conversational Speech Corpus

    Spontaneous, real-world conversational speech in four rare Indian languages — transcribed, segmented, and ready for ASR and speech-LLM training.

    Languages:
    Santali, Gondi, Tulu, Bhojpuri
    Size:
    260 hrs · 1,850 speakers
    IAA:
    0.85

    Licence from

    Contact us

    Rights-cleared
    View data card See pricing

Browse datasets by language

Dedicated pages for each language — including the low-resource Indic and African languages most vendors can't reach.

Browse datasets by domain

Vertical landing pages for each domain — built for the way each industry actually talks.

Languages we cover, across every continent

Our coverage spans a 400-language footprint — Indic and low-resource languages first, extending across Africa, Asia, the Americas and beyond. Select a continent to filter the list, or search by name, native script or ISO code. Multilingual data infrastructure built for language sovereignty, not just the major markets.

India is shown as a map because it is our deepest area of language coverage. Below it, a grid of continents is shaded by how many languages we cover; select a continent to filter the language list. A full text list of every language, grouped by continent, follows as the accessible equivalent of this map.
Accurate map of India with correct official national boundaries — all states and union territories including Jammu & Kashmir and Ladakh, shaded by region, with disputed and claimed areas hatched per the Government of India. India is our deepest area of language coverage.

India — deepest Indic & low-resource coverage

South Asia flagged — our deepest Indic & low-resource coverage

India shown as a map; continents as a coverage grid — see the full searchable list (right) for every language we cover.

Continent:

languages shown (filtered) · 400-language footprint in scope

Language coverage

Which languages does Cognegica's dataset registry cover?

Cognegica enumerates 392 languages across the registry within a 400-language collection footprint, with deep coverage of the low-resource long tail. Counts below are drawn directly from the registry, grouped by continent.

Region / continentLanguages enumeratedLow-resource
South Asia11097
Asia4335
Middle East119
Central Asia108
Africa132131
Europe1811
Americas98
Oceania66
Global / Other5337
Total enumerated registry392342

Enumerated registry counts by continent. The headline 400-language footprint includes additional collectable languages scoped per project.

FAQ

Dataset catalog — common questions

What kind of datasets do you license?

Proprietary, rights-cleared, sovereign datasets for low-resource and Indic languages across a 400-language footprint — covering AI safety and evaluation, legal, healthcare, general multilingual, and Physical AI domains, in text, audio, speech, image, video and sensor modalities.

Are the datasets rights-cleared and safe to train on?

Yes. Every dataset is authored or collected from consenting native speakers under agreements that grant commercial reuse, with documented provenance and consent per record — never scraped. Published quality metrics include inter-annotator agreement and QA pass rates.

How does dataset licensing work?

Each dataset offers commercial non-exclusive and enterprise/sovereign exclusive licensing. Licensing is scoped and quoted per dataset against languages, modalities, volume and rights. Contact us for terms.

Can the data stay in India (data residency)?

Yes. We offer an India data-residency / sovereign delivery option across collection, processing, annotation and storage, aligned with the DPDP Act, 2023. See Sovereign Data.

Do you have Physical AI datasets?

Yes — Physical AI is a data service, not research. Browse multi-sensor and annotation samples at Physical AI datasets, or engage our collection and annotation services to build to spec.

Can you build a custom dataset in a language not listed?

Yes. When the data you need does not exist, we collect it to spec across our 400-language footprint. Request a custom dataset.

Don't see the language or domain you need?

We build proprietary, rights-cleared datasets to spec — across 400 Indian languages and beyond. Tell us your target languages, modality, and quality bar.

Request a custom dataset → Talk to us