TTS Eval Datasets: Zero-Shot Text-to-Speech Benchmarks

Multiple test sets for evaluating zero-shot text-to-speech models, introduced in associated research papers. The collection includes Chinese and English dialogue testsets, an English LibriSpeech testset, and a 24-language multilingual testset. The repository is maintained by k2-fsa and was last updated in April 2026.

Use Cases

• Benchmarking zero-shot TTS model performance based on the provided multilingual test sets.
• Evaluating spoken dialogue generation models based on the Chinese and English dialogue testset.
• Comparing TTS model outputs across languages based on the 24-language multilingual testset.

Strengths

• Includes a 24-language multilingual testset for broad linguistic evaluation.
• Test sets are directly linked to specific, named research papers (ZipVoice-Dialog, F5-TTS, Seed-TTS).
• Last updated on 2026-04-22, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.
• License information is unknown, which may restrict usage.

Provenance

Source

k2-fsa on Hugging Face.

Collection Method

Likely curated for research and benchmarking purposes, as each testset is introduced in a specific paper.

Freshness

Last updated 2026-04-22 05:05:30.

Geography

Multilingual, with specific testsets for Chinese and English.