This dataset contains augmented speech recordings and transcriptions for isiXhosa, one of South Africa's official languages.
The dataset has been optimized for use with OpenAI's Whisper ASR models.
Dataset Statistics
Number of samples: 680
Language: isiXhosa (Xho)
Audio format: WAV, 16kHz, mono, 16-bit
Maximum duration: 30 seconds (truncated for Whisper compatibility)
Transcription format:… See the full description on the dataset page: https://huggingface.co/datasets/zionia/isixhosa-asr-train.
