Multilingual Synthetic TTS Dataset with 68,677 Clips Across 9 Languages

68,677 synthetic speech clips across 9 languages, generated using the Qwen3-TTS-12Hz-1.7B-Base model with zero-shot voice cloning from 5 reference speakers. The dataset was submitted to the Uncharted Data Challenge hosted by Adaption Labs and is authored by Reubencf. It was last updated on 2026-04-15.

Use Cases

• Training text-to-speech models based on multilingual synthetic audio.
• Evaluating voice conversion systems based on zero-shot voice cloning from reference speakers.
• Benchmarking automatic speech recognition models on synthetic multilingual speech.
• Developing multilingual speech synthesis models based on a dataset covering 9 languages.

Strengths

• Contains 68,677 audio clips, providing a substantial volume of data.
• Covers 9 languages, offering multilingual scope.
• Generated using a specific, named model (Qwen3-TTS-12Hz-1.7B-Base) with a defined method (zero-shot voice cloning from 5 speakers).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the synthetic generation model and source languages.

Provenance

Source

Reubencf on Hugging Face.

Collection Method

Synthetically generated using Qwen3-TTS-12Hz-1.7B-Base model with zero-shot voice cloning.

Freshness

Last updated 2026-04-15 10:02:44; freshness should be verified.