Dutch TTS Dataset - Complete Labeled
A comprehensive Dutch text-to-speech dataset with 596,508 audio samples totaling 234GB of audio data.
Quick Preview
The default config shows a 100-row sample for the dataset viewer. To access the full dataset, use the full config.
Dataset Description
This dataset contains Dutch speech recordings with rich metadata including:
- Emotion labels (neutral, happy, sad, angry)
- Speaker IDs (239,388 unique speakers)
- Prosodic features (pitch mean/std, speaking rate)
- Audio quality metrics (SNR)
- Normalized text transcriptions
Dataset Statistics
| Statistic | Value |
|---|---|
| Total samples | 596,508 |
| Total audio size | ~234 GB |
| Audio format | WAV, 16kHz mono |
| Unique speakers | 239,388 |
| Avg duration | ~10 seconds |
Emotion Distribution
- neutral: 545,079 (91.4%)
- happy: 51,124 (8.6%)
- sad: 291 (0.05%)
- angry: 40 (0.01%)
Loading the Dataset
from datasets import load_dataset
# Load sample (100 rows) - for quick testing
dataset = load_dataset("AITRADER/dutch-tts-labeled-complete")
# Load full dataset (596k rows) - use streaming for large data
dataset = load_dataset("AITRADER/dutch-tts-labeled-complete", "full", streaming=True)
for sample in dataset["train"]:
print(sample["text"], sample["emotion"])
# sample["audio"] contains the audio array
break
Features
| Feature | Type | Description |
|---|---|---|
| audio | Audio | Audio waveform (16kHz) |
| sampling_rate | int | Always 16000 Hz |
| duration | float | Duration in seconds |
| text | string | Original transcription |
| text_normalized | string | Normalized transcription |
| speaker_id | string | Unique speaker identifier |
| emotion | string | Emotion label |
| emotion_confidence | float | Confidence score (0-1) |
| valence | float | Emotional valence (0-1) |
| arousal | float | Emotional arousal (0-1) |
| pitch_mean | float | Mean pitch in Hz |
| pitch_std | float | Pitch standard deviation |
| words_per_minute | float | Speaking rate |
| snr_db | float | Signal-to-noise ratio in dB |
| dataset_source | string | Original dataset source |
Data Sources
- facebook/multilingual_librispeech (60%)
- freds0/cml_tts_dataset_dutch (40%)
- google/fleurs (0.1%)
License
CC-BY-4.0