Lahgtna Levantine TTS: 50,000 Synthetic Speech Utterances

A synthetic speech dataset contains 50,000 utterances of Levantine Arabic dialect and Arabic-English code-switched speech. It was generated using the Lahgtna-OmniVoice fine-tuned TTS model and includes 10 speakers. The dataset page was last updated on 2026-05-30.

Use Cases

• Train Levantine Arabic TTS models based on the 44,154 pure dialect utterances.
• Develop code-switching speech synthesis systems based on the 5,846 mixed-language utterances.
• Benchmark multi-speaker TTS performance based on the 10-speaker (5 male, 5 female) structure.
• Create synthetic speech assets for applications targeting Levantine Arabic speakers.

Strengths

• Contains 50,000 total utterances, providing a substantial corpus for model training.
• Includes a balanced speaker set of 10 individuals (5 male, 5 female).
• Offers a specific focus on Levantine Arabic dialect with 44,154 pure dialect samples.
• Provides a dedicated subset of 5,846 utterances for Arabic-English code-switching research.

Limitations

• Data is synthetically generated, which may not fully capture natural speech characteristics.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Synthetically generated using the Lahgtna-OmniVoice fine-tuned zero-shot TTS model.

Freshness

Last updated 2026-05-30 14:36:46; freshness should be verified.

Geography

Likely covers the Levantine Arabic dialect region.