Persian Speech Suite: GE2PE & TTS
A unified playground for Persian grapheme‑to‑phoneme conversion (GE2PE) and text‑to‑speech synthesis (Mana TTS).
✨ Now supports long texts! The TTS system automatically splits long texts into natural segments. And also converts numbers to Persian text for better pronunciation.
Convert Persian text to its phonemic transcription. Choose between Homo‑GE2PE and Homo‑T5, optionally applying short‑vowel rules and/or a custom dictionary.
Examples
Generate natural‑sounding Persian speech from your text using Tacotron2 + HiFiGAN.
✨ New features:
- Long text support: Automatically splits text into natural segments with optional pauses
- Smart number conversion: Numbers (۱۴۰۲, 2025, ۵۰۰۰) are automatically converted to text
Adds 300ms pause between text segments for natural flow
Examples
Acknowledgments
- Nasl‑e‑Mana, the monthly magazine of the blind community of Iran
- ManaTTS Dataset
- Persian‑MultiSpeaker‑Tacotron2
- Homo-GE2PE (Github)
- Base GE2PE Paper
- Base GE2PE Model
- HomoRich Dataset (Huggingface)
- HomoRich Dataset (Github)
- SentenceBench Persian G2P Benchmark
Citation
@misc{qharabagh2025fastfancyrethinkingg2p,
title={Fast, Not Fancy: Rethinking G2P with Rich Data and Rule-Based Models},
author={Mahta Fetrat Qharabagh and Zahra Dehghanian and Hamid R. Rabiee},
year={2025},
eprint={2505.12973},
archivePrefix={arXiv},
primaryClass={cs.CL},
}
@article{fetrat2024manatts,
title={ManaTTS Persian: A Recipe for Creating TTS Datasets for Lower-Resource Languages},
author={Mahta Fetrat Qharabagh and Zahra Dehghanian and Hamid R. Rabiee},
journal={arXiv preprint arXiv:2409.07259},
year={2024},
}