Research Projects Blog Agent Skill Publications Contact
Projects  /  Saudi Arabic TTS (KSA)

🗣  Saudi Arabic TTS (KSA)

Qwen3-based TTS model fine-tuned for Saudi Arabian Arabic dialect with dialect-specific text normalization and production-grade speech synthesis.

TTS
Arabic
Saudi
KSA
Qwen3

Overview

The Saudi Arabic TTS (KSA) project brings high-quality speech synthesis to the Saudi Arabian Arabic dialect — building on the learnings and infrastructure from the Emirati Arabic TTS project. Saudi Arabic presents its own phonological and prosodic characteristics distinct from other Gulf dialects, requiring dialect-specific training data and normalization rules.

Model

qwen3-TTS-KSA

LLM-based TTS model fine-tuned on Qwen3 architecture for Saudi Arabic (KSA) dialect synthesis. Leverages a large language model’s deep understanding of Arabic morphology and phonology to produce more natural-sounding speech than classical acoustic pipelines.

  • Base model: Qwen3 (fine-tuned for TTS)
  • Language: Saudi Arabian Arabic dialect
  • Approach: LLM-based speech synthesis
  • Published on HuggingFace

Dialect Coverage

Together with the Emirati Arabic TTS project, this work provides TTS coverage for two major Gulf Arabic dialects:

  • Emirati Arabic — UAE dialect, bilingual Arabic/English support
  • Saudi Arabic (KSA) — Saudi dialect, dedicated dialect-specific model

Text Normalization

The KSA TTS system includes an extended text normalization pipeline adapted for Saudi Arabic conventions:

  • Saudi numeral and currency verbalization
  • Dialect-specific abbreviation expansion
  • Arabic G2P adapted for KSA phonological features
  • Mixed Arabic/English codeswitching handling