Catalogue

Explore African languages

languages with data good enough to publish

countries where one of them is the language of record

countries reached once cross- border speakers are counted

Africa has somewhere between 1,250 and 2,100 natively spoken languages depending on where the line between language and dialect is drawn. This catalogue holds the ones where public data is good enough to say something true — West, East, Southern, Central and the Horn on the same terms — and stops there. Building the other two thousand rows from guesses would make the whole thing worse than useless.

Resource tierCountryFamily

Showing 29 of 29 languages

No filters applied

29 African languages with their country, resource tier, tone, dialect count, AI readiness score and speaker estimate.
Language	Country	Tier	Tone	Dialects	AI readiness	Speakers
AfrikaansafrThe best-served language on the continent, and the control case: where a language is genuinely covered, the only place left to compete is price and domain.	CountrySouth Africa	TierServed	ToneNon-tonal	Dialects2 dialects	AI readiness100	Speakers17m
SwahiliswhThe least attractive large African language commercially, precisely because it is the one the majors already serve tolerably — ElevenLabs places it in its 5–10% WER band, the only African language above Moderate.	CountryTanzania	TierServed	ToneNon-tonal	Dialects4 dialects	AI readiness91	Speakers80m
Nigerian Englisheng-NGWhere the enterprise volume actually sits. Frontier ASR runs 24.7–35% WER on the accent generally — and 40–70% on African named entities, which is the number that decides whether a bank can deploy.	CountryNigeria	TierServed	ToneNon-tonal	Dialects2 dialects	AI readiness76	Speakers110m
HausahauThe best cross-border expansion asset in West Africa — one language investment reaches Niger, Chad, Ghana and northern Cameroon. Also the one West African language ElevenLabs does cover, so the deficit here is smaller than it looks.	CountryNigeria	TierLow-resource	ToneTonal	Dialects5 dialects	AI readiness65	Speakers88m
KinyarwandakinThe clearest proof that community collection outperforms top-down prioritisation: 1,211 hours onto Common Voice from 420 contributors, which no desk ranking would have predicted.	CountryRwanda	TierLow-resource	ToneTonal	Dialects1 dialect	AI readiness65	Speakers15m
SomalisomA large diaspora is the unusual feature: remittance and telehealth demand sits partly outside the continent.	CountrySomalia	TierLow-resource	ToneTonal	Dialects3 dialects	AI readiness59	Speakers22m
YorubayorLagos-concentrated, so co-located with the buyers. Absent from ElevenLabs TTS, AWS Transcribe and Google Chirp 3; ElevenLabs' own published STT band for it is 25–50% WER.	CountryNigeria	TierLow-resource	ToneTonal	Dialects7 dialects	AI readiness49	Speakers46m
IgboiboCommercially dense in trade and diaspora remittances. Absent from AWS Transcribe entirely and capped at Chirp 2 on Google.	CountryNigeria	TierLow-resource	ToneTonal	Dialects6 dialects	AI readiness49	Speakers33m
AmharicamhA non-Latin script, so the text pipeline is a different problem from every other language on this list. Big speaker base, difficult FX and regulatory environment.	CountryEthiopia	TierLow-resource	ToneNon-tonal	Dialects3 dialects	AI readiness49	Speakers57m
ZuluzulThe largest free corpus covering it carries a CC BY 4.0 headline and still prohibits text-to-speech and voice cloning — which supports an ASR entry and not a TTS one.	CountrySouth Africa	TierLow-resource	ToneTonal	Dialects2 dialects	AI readiness49	Speakers28m
LugandalugUnusually well served by open research relative to its size, largely through one sustained local effort.	CountryUganda	TierLow-resource	ToneTonal	Dialects1 dialect	AI readiness49	Speakers11m
XhosaxhoClick consonants make it acoustically distinctive, which is a modelling cost and a differentiator at once.	CountrySouth Africa	TierLow-resource	ToneTonal	Dialects3 dialects	AI readiness43	Speakers19m
Nigerian PidginpcmThe lingua franca of urban informal commerce from Lagos to Accra to Douala, supported by no major vendor at all, with no tonal orthography problem. The single best first language for a West African voice product.	CountryNigeria	TierLow-resource	ToneNon-tonal	Dialects4 dialects	AI readiness35	Speakers80m ±
TwitwiGhana has the strongest existing developer community for its own languages on the continent, which makes it the cheapest second market to enter.	CountryGhana	TierLow-resource	ToneTonal	Dialects3 dialects	AI readiness35	Speakers17m
OromoormThe largest African language with effectively no commercial speech support anywhere. Its Latin orthography was adopted in 1991, so the text pipeline is cheap — which makes the absence of coverage harder to explain than for Amharic next door.	CountryEthiopia	TierExtremely low-resource	ToneNon-tonal	Dialects4 dialects	AI readiness35	Speakers37m
ShonasnaSits in ElevenLabs' own Moderate band at 25–50% word error rate, which is published evidence that its size has not bought it coverage. A large diaspora in South Africa and the UK puts part of the demand outside Zimbabwe.	CountryZimbabwe	TierLow-resource	ToneTonal	Dialects4 dialects	AI readiness35	Speakers15m
Northern SothonsoAn official language of South Africa, which means a statutory obligation to serve it and a public procurement route that does not wait on commercial demand. Covered by the NCHLT corpora and by nothing any vendor ships.	CountrySouth Africa	TierLow-resource	ToneTonal	Dialects3 dialects	AI readiness35	Speakers4.7m
WolofwolFrancophone West Africa's entry point, and the place where French–Wolof code-switching is the norm rather than the exception.	CountrySenegal	TierLow-resource	ToneNon-tonal	Dialects3 dialects	AI readiness31	Speakers12m
LingalalinKinshasa is one of the largest cities on the continent and francophone Central Africa is served by essentially nobody. The pivot language here is French, which is a different pipeline from every English-pivot row above it.	CountryDR Congo	TierExtremely low-resource	ToneTonal	Dialects3 dialects	AI readiness31	Speakers40m ±
BambarabamThe Sahel's main lingua franca, and the clearest case where humanitarian and agricultural demand arrives long before commercial demand. Mande rather than Bantu or Volta-Niger, so none of the transfer learning elsewhere in this catalogue reaches it.	CountryMali	TierExtremely low-resource	ToneTonal	Dialects3 dialects	AI readiness31	Speakers15m ±
FulfuldefulThe highest-value regional asset after Hausa: cross-border into Niger, Cameroon and Chad, with strong humanitarian and agricultural demand.	CountryNigeria	TierExtremely low-resource	ToneNon-tonal	Dialects4 dialects	AI readiness18	Speakers20m ±
IbibioibbState-government and healthcare procurement is the realistic route in. No vendor covers it and no open corpus of scale exists.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects2 dialects	AI readiness5	Speakers11m ±
KanurikncNorth-east Nigeria and the Lake Chad basin, where NGO and humanitarian demand is real and durable even where commercial demand is not.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects3 dialects	AI readiness5	Speakers8.2m ±
TivtivNamed in Nigeria's national language-model programme, which makes it a candidate for publicly funded collection ahead of commercial demand.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects2 dialects	AI readiness5	Speakers4.6m ±
IjawijcNiger Delta. Oil-sector and state-government relevance out of proportion to the speaker count, and a cluster rather than one language.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects4 dialects	AI readiness5	Speakers2.4m ±
EfikefiA long orthographic and lexicographic tradition relative to its size, which makes text collection unusually cheap here.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects1 dialect	AI readiness5	Speakers2.0m ±
EdobinBenin City and the Edo heartland. Cultural-heritage funding is a more realistic first buyer than any enterprise.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects1 dialect	AI readiness5	Speakers1.5m ±
NupenupMiddle Belt. Collection here needs a named funder — a state contract or a research grant — before it is worth starting.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects2 dialects	AI readiness5	Speakers1.2m ±
IgalaiglClosely related to Yoruba, which makes it the clearest test of whether transfer learning inside a subgroup earns its keep.	CountryNigeria	TierExtremely low-resource	ToneTonal	Dialects1 dialect	AI readiness5	Speakers1.8m ±

Africa has upwards of a thousand more languages than this. Public data good enough to publish a row about exists for a small fraction of them — this catalogue stops where the sources do rather than filling the rest in.

How the tiers work

Resource tiers

Where a language sits decides what is worth building for it

The split is keyed to the 200-hour threshold at which speech recognition becomes production-usable. Past a few hundred hours per language, label quality rather than volume is the binding constraint — in one published study 38.6% of high-error cases traced to noisy ground truth rather than insufficient data.

Served3: 200+ hours of transcribed speech, commercial voices, big-tech support. Compete on domain, not capability.
Low-resource14: Some open data and big-tech ASR, but no production TTS. Demand exists; quality does not.
Extremely low-resource12: Under ten hours, benchmark-only presence. Collection required before anything ships.
Undocumented0: No digital text or audio at all. Preservation work, not a product.

Missing a language you speak?

The catalogue grows from contributor demand rather than from a desk estimate. A language whose speakers organise to record it is worth more than a bigger number with no community behind it.

Become a contributor See the research