A gold-standard evaluation set of 1,072 speech samples annotated with 57 voice taxonomy dimensions on a 0-6 ordinal scale by the Gemini 3.1 Pro model. The dataset was created by TTS-AGI and last updated on April 8, -2026. It is part of a related series for pre-training and fine-tuning voice attribute classifiers.
Use Cases
- Benchmarking voice attribute classifiers based on the 57 annotated dimensions.
- Evaluating the performance of AI models on ordinal-scale voice taxonomy tasks.
- Training or fine-tuning models for speech feature recognition using the provided gold-standard labels.
- Conducting research on voice characteristics and their automated measurement.
Strengths
- Contains 1,072 annotated speech samples, providing a substantial evaluation corpus.
- Features 57 distinct voice taxonomy dimensions, suggesting a detailed annotation schema.
- Serves as a designated gold-standard evaluation set for model benchmarking.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- TTS-AGI via Hugging Face.
- Collection Method
- Speech samples annotated by the Gemini 3.1 Pro AI model.
- Time Range
- null
- Freshness
- Last updated 2026-04 08 07:09:29; freshness should be verified.
- Geography
- null