Vocal Burst Annotation ASR Tuning Dataset: 500,000 Synthetic Multilingual Audio Samples

500,000 synthetic audio samples for training automatic speech recognition models. The dataset, created by TTS-AGI, contains approximately one minute of audio per sample with interleaved speech and vocal bursts. It was last updated on Hugging Face in April 2026.

Use Cases

• Training ASR models to transcribe speech alongside non-verbal vocalizations based on the described inline vocal burst captioning.
• Fine-tuning speaker diarization systems using the described speaker and timestamp annotations.
• Developing audio event detection models for vocal bursts like laughs and coughs based on the described annotation scheme.
• Creating synthetic training data for multilingual speech processing pipelines based on the described multilingual nature.

Strengths

• 500,000 samples provide a substantial volume for model training.
• Each sample contains precise timing information for speech segments and vocal bursts as described.
• The dataset is multilingual, which likely increases its applicability across languages.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is synthetic, which may not perfectly replicate real-world audio distributions.

Provenance

Source

TTS-AGI via Hugging Face.

Collection Method

Synthetically generated.

Freshness

Last updated 2026-04-11 19:31:23; freshness should be verified.