Logo Lanfrica

AfriSpeech Afri-10

Domaine:

natural language processing

Type de record:

dataset
Créateur:
Afr
Hôte:
A compact, balanced multilingual speech corpus covering 10 African languages, with roughly 10 hours of clean read speech per language (~100 hours total) and aligned transcripts. It is designed as a ready-to-use starting point for text-to-speech (TTS) and automatic speech recognition (ASR) experiments across a diverse set of African languages and writing systems. The selection was curated with the afrispeech-selector tool (top 10