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.

Downloads last month
36
Safetensors
Model size
0.2B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support