Magpie TTS Arabic
Saudi & Emirati Dialect Speech Synthesis
🏆 60k+ Downloads on Hugging Face
Dialect-adapted multilingual speech synthesis adapting NVIDIA's Magpie architecture to conversational Gulf Arabic.
What it does
- Fine-tuned NVIDIA Magpie TTS Multilingual 357M on conversational speech, adapting the architecture to ar-SA and ar-AE dialectal phonetics and non-standard orthography.
- Engineered an end-to-end speech data curation pipeline: 22.05 kHz resampling, noise filtering, Silero-VAD utterance segmentation, and speaker diarization.
- Published open-source model checkpoints, curated datasets, and interactive Hugging Face Spaces demos with over 60,000+ community downloads.