RGAD Cross-Lingual TTS: 10-Hour Prompt-Conditioned Chinese Speech Dataset

RGAD Cross-Lingual TTS 10h is a 10-hour speech dataset for prompt-conditioned text-to-speech fine-tuning. It was created by author 'isabeth' and last updated on Hugging Face in May 2026. The dataset contains paired audio prompts and targets across languages, specifically for generating Chinese speech from prompts in other languages.

Use Cases

• Fine-tuning TTS models for Chinese speech generation based on cross-lingual prompts.
• Training voice conversion systems using paired prompt and target audio files.
• Researching cross-lingual speech synthesis based on the described prompt-target language pairs.
• Developing prompt-conditioned TTS systems based on the structured JSONL metadata.

Strengths

• Dataset contains 10 hours of audio data, providing a substantial resource for model training.
• Data is structured with explicit language tags (e.g., 'en-US', 'zh-CN') for cross-lingual tasks.
• Includes paired prompt and target audio files, which are essential for supervised learning in TTS.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Hugging Face dataset uploaded by user 'isabeth'.

Collection Method

Likely collected or synthesized for TTS fine-tuning, but specific gathering method is not described.

Freshness

Last updated 2026-05-28 13:57:07; freshness should be verified.