Fahd1199/darija-tts-vits-ready

Darija TTS Dataset (VITS-Ready)

This is a processed version of the Fahd1199/darija-tts dataset, formatted specifically for finetuning VITS/MMS TTS models.

Dataset Structure

The dataset contains:

• 1740 training samples
• 436 validation samples

Each sample has:

• audio: Path to a 16kHz WAV file
• text: Darija (Moroccan Arabic) transcription
• speaker_id: Speaker identifier (single speaker)

How to Use

This dataset is ready to be used with the finetune-hf-vits repository for finetuning MMS TTS models.

Example config:

{
    "model_name_or_path": "facebook/mms-tts-ara",
    "dataset_name": "Fahd1199/darija-tts-vits-ready",
    "text_column_name": "text",
    "audio_column_name": "audio",
    "speaker_id_column_name": "speaker_id"
}