metadata
license: apache-2.0
language:
- yo
- ha
- ig
- en
tags:
- automatic-speech-recognition
- asr
- whisper
- nigerian
- yoruba
- hausa
- igbo
- multilingual
Nigerian Multilingual ASR
An automatic speech recognition (ASR) model that transcribes spoken audio in four Nigerian languages.
Supported Languages
| Language | Code |
|---|---|
| English | en |
| Yoruba | yo |
| Hausa | ha |
| Igbo | ig |
What it does
This model converts spoken audio to text. It accurately transcribes speech in Yoruba, Hausa, Igbo, and Nigerian-accented English. It handles natural conversational speech and is optimised for Nigerian speakers.
Usage
Send audio files to the transcription endpoint. Pass the language code for best accuracy — the model also auto-detects language when no code is provided.
Training Data
Fine-tuned on a multilingual dataset of Nigerian speech covering Yoruba, Hausa, Igbo, and English, with a combined total of hundreds of thousands of sentences across all four languages.