Nabra Saudi

Saudi-dialect Arabic TTS: an 82M voice and the 7.48M distilled from it. Same front end, same voice pack, 24 kHz — the only variable is the model. Type dialect text the way it is normally written, without full tashkeel.

Both were trained on 146 hours of single-speaker Saudi speech; the small model is a distilled copy of the large one, not a separately trained model.