For enterprise

Voice AI that works in the languages your customers actually speak.

ASR, TTS, and call-centre intelligence APIs for Nigerian English, Pidgin, Yoruba, Hausa, and Igbo — built on ethically-sourced cooperative data with a documented consent chain.

Why existing solutions fail in Nigeria

Whisper and Google fail on Nigerian indigenous languages

Whisper large-v3 produces approximately 55% word error rate on Igbo. Google Speech-to-Text does not support Yoruba or Hausa. AWS Transcribe does not support any Nigerian indigenous language. For voice AI products serving Nigerian users, this is the difference between usable and unusable.

Existing African language datasets cannot be used commercially

Most African language speech data is research-tier — collected for academic use, with no documented contributor consent for commercial AI training. Enterprise legal teams cannot approve them. Your AI ethics audit will flag them. We built the consent chain that makes commercial use possible.

There is no domain-specific data for the use cases that matter

A bank needs financial-services speech data. A telco needs customer-service dialogue. A health programme needs PHC interaction. Generic speech corpora do not deliver this. Africa's Voice collects against ten domains per language, with explicit metadata tagging.

What we offer

ASR API

From $0.003 / audio minute

Real-time and async transcription endpoints. Five Nigerian languages at launch. SFTP drop integration available alongside REST. SLA-backed for enterprise pilots.

✓5 Nigerian languages at launch
✓Real-time + async endpoints
✓Code-switching tolerant
✓SFTP drop integration available

Datasets

From $5,000 per dataset

Validated speech corpora with paired human-verified transcripts, domain tags, and full consent provenance. Research tier for academic use; commercial tier for AI training.

✓Audio + paired transcripts
✓Domain-tagged metadata
✓NDPA 2023 compliant
✓Tiered: research / commercial

Enterprise CCI

From $2,000 / month

Call Centre Intelligence: 100% transcription, sentiment scoring, compliance flagging. Designed for CBN-regulated banks and telcos with disclosure requirements.

✓Full call transcription
✓Compliance flagging
✓Multi-language quality scoring
✓On-premise deployment available

The differentiator

The consent and governance story

Every audio file in our corpus has a documented consent record. Contributors are members of the Africa's Voice Data Cooperative. They consented at one of three tiers — research only, commercial use, or public domain. They can withdraw consent at any time. The audit chain is queryable per file. This is what your AI ethics audit needs to see — and what most African language datasets cannot provide.

API preview

POST /v0/transcribe
Authorization: Bearer {api_key}
Content-Type: audio/wav

→  Response:
{
  "language_detected": "ibo",
  "language_confidence": 0.94,
  "transcript": "Kọọ ihe i mere n'oge ụtụtụ taa.",
  "duration_seconds": 5.2,
  "model_version": "av-asr-v0.1"
}

Discovery call

Start a discovery conversation

20 minutes. We will not pitch — we want to understand your problem first. Reply within 48 business hours.