Igbo AI

Igbo AI that speaks
from the south-east.

The only commercial AI voice platform with native Igbo speech recognition and text-to-speech — subdot diacritics (ị, ụ, ọ) fully preserved, Igbo–English code-switching handled naturally. Built for Enugu, Onitsha, Owerri, and Port Harcourt businesses.

Request beta See Igbo pricingHear Maraba in Igbo →
45M+
Igbo speakers in Nigeria
3
subdot diacritics fully supported
<800ms
language detection latency
24/7
answers in Igbo, always

What Maraba's Igbo AI actually does

Most voice AI platforms either skip Igbo entirely or produce broken text by stripping the subdot characters that are fundamental to the language. Maraba was built from the ground up to handle Igbo correctly.

🎙️
Igbo Speech Recognition (STT)
Maraba transcribes spoken Igbo with subdot diacritics preserved — ị, ụ, ọ appear correctly in every transcript. Enugu, Onitsha, and Owerri accent variants are all supported.
🔊
Igbo Text-to-Speech (TTS)
Maraba speaks back in natural Igbo — using subdot-aware pronunciation to produce speech that native speakers recognise as authentic, not foreign-accented machine output.
🔀
Igbo–English Code-Switching
South-east Nigerian callers switch between Igbo and English constantly in commercial conversations. "Ọ dị mma, but what's the delivery cost to Owerri?" — Maraba handles this without hesitation.
🔄
Real-time Language Detection
Maraba identifies Igbo speech within 800ms of call start and responds immediately — no IVR menus, no "please select your language", no friction for callers.
✍️
Subdot-Safe Transcripts
We never corrupt Igbo by stripping ị → i, ụ → u, or ọ → o. These are phonetically distinct sounds. Your call records are accurate and usable for downstream systems.
📋
English Summaries Delivered to WhatsApp
Every Igbo call produces an English summary — intent, key details, action items — sent to your WhatsApp within 60 seconds. Your whole team can act without needing to speak Igbo.

Igbo subdot diacritics — handled correctly

Igbo's three subdot letters represent distinct vowel sounds that other AI platforms routinely destroy. Maraba preserves all three.

open front vowel — ịchọ (to find)
back unrounded — ụlọ (house)
open-mid back — ọ dị mma (it is good)
uppercase variant preserved
uppercase variant preserved

When a caller says "Achọrọ m ịghọta ụlọ ọrịa ebe a" (I want to know the hospital location here), the transcript you receive preserves every subdot character exactly. This is not a minor detail — it is the difference between a usable record and a corrupt one.

How a real Igbo call sounds

This example shows Maraba handling a call from an Onitsha trader asking about bulk order availability — mixing Igbo and English as is typical in south-east business conversations.

Live call example — Onitsha wholesale trader

Caller: Ndeewo! Achọrọ m ịghọta ma ụlọ ahịa unu nwere akwa ọchịchọ anọ.

Maraba: Ndeewo! Ee, anyị nwere akwa ọchịchọ anọ. Ị chọrọ àgwà maka ụdị kạ ọ dị? Anyị nwere cotton, polyester, na lace.

Caller: I need cotton — how much for bulk, like 50 yards?

Maraba: For 50 yards of cotton, the bulk rate is ₦1,800 per yard — that's ₦90,000 total. Delivery to Onitsha is included for orders above ₦80,000. Can I note your name for a reservation?

Caller: Yes — Chidinma Okonkwo.

Maraba: Daalu, Chidinma. I've logged your interest for 50 yards of cotton. The team will call you back to confirm stock and delivery timing.

Integrate Igbo STT via API

Developers can submit Igbo audio to the Maraba REST API and receive subdot-accurate transcripts. Header-only authentication — no query params.

# POST audio for Igbo transcription
import requests

response = requests.post(
    "https://maraba.ai/api/v1/transcribe/",
    headers={
        "X-API-Key": "your_api_key_here",
        "Accept": "application/json",
    },
    json={
        "audio_url": "https://storage.example.com/call-igbo-001.wav",
        "language_hint": "ig",  # ISO 639-1 — Igbo
        "preserve_diacritics": True,  # critical for ị, ụ, ọ
    }
)

transcript = response.json()
# {
#   "text": "Ndeewo! Achọrọ m ịghọta ụlọ ahịa unu",
#   "language_detected": "ig",
#   "confidence": 0.96,
#   "diacritics_preserved": true
# }

Built for south-east Nigerian businesses

Every business in Igboland that fields inbound calls in Igbo can benefit — but these verticals account for the highest call volumes.

🏪 Onitsha wholesale and retail traders
Onitsha market generates enormous Igbo call traffic — bulk pricing queries, stock availability checks, delivery confirmations. Maraba handles them all without tying up your sales team's time.
🏥 Clinics (Enugu, Owerri, Awka)
Appointment booking, doctor availability, and emergency triage in Igbo. Maraba answers at 6am, on weekends, and during the 12–2pm lunch rush when your receptionist is unavailable.
🚢 Logistics (Port Harcourt, Aba corridor)
Cargo tracking, delivery ETAs, and rate enquiries for the Port Harcourt–Aba–Onitsha corridor are common Igbo call types. Maraba gives accurate answers from your knowledge base in real time.
💊 Pharmacies (Enugu, Umuahia, Nnewi)
Patients call in Igbo to check medication stock, confirm dosages, and ask about branch hours. Maraba handles these calls correctly — never guessing, always drawing from your configured data.

Where Maraba serves Igbo-speaking businesses

Available on all Nigerian networks across every south-east and south-south state.

Enugu Onitsha Owerri Port Harcourt Awka Abakaliki Umuahia Asaba Nnewi Aba Nsukka Orlu

Frequently asked questions

Common questions from south-east Nigerian business owners about Maraba's Igbo AI voice platform.

Does Maraba preserve Igbo subdot diacritics like ị, ụ, and ọ?
Yes — always. Stripping Igbo subdot characters produces corrupt text. ụlọ (house) becomes ulo, which means nothing. Maraba never strips diacritics from Igbo text in any pipeline stage — transcription, storage, or TTS input.
Can Maraba handle the Igbo–English switching common in Onitsha business calls?
Yes — this is a core design requirement, not a workaround. Igbo–English code-switching is extremely common in commercial south-east Nigerian conversations. Maraba tracks both languages simultaneously and responds naturally in whichever the caller is currently using.
Which Igbo dialect does Maraba support?
Maraba is trained on Central Igbo — the standardised dialect used in education and broadcasting — which is widely understood across all Igbo-speaking states. The model also handles common pronunciation and vocabulary variations from Enugu, Onitsha, Owerri, and Port Harcourt.
Is Maraba suitable for Onitsha market traders with high Igbo call volumes?
Yes — Onitsha market traders are among our most active users. Maraba handles bulk pricing enquiries, stock availability checks, and delivery confirmations in Igbo 24/7 — freeing traders to close deals instead of answering repetitive calls.
Is there an Igbo STT API for developers?
Yes. The Maraba API at /api/v1/ accepts audio with language_hint "ig" for Igbo. Authentication uses X-API-Key header only — no query parameters. Documentation is in the developer portal at maraba.ai/developer/.

Ọ dị mma — bido taa.
Apply for beta access.

Set up Maraba in under 10 minutes. No hardware. No Igbo-speaking staff required around the clock. Every caller deserves a real answer.

Request beta →    Hear Maraba in Igbo