Datasets:

zakihassan
/

som_tts

Tasks:

Text-to-Speech

Modalities:

Formats:

Languages:

Size:

Libraries:

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Viewer

Auto-converted to Parquet Duplicate

Split (1)

train · 9.88k rows

Search is not available for this dataset

audio audioduration (s) 0.5 19.7

End of preview. Expand in Data Studio

Somali TTS Dataset

This dataset contains high-quality Somali speech recordings with corresponding transcriptions.
It is designed for training Text-to-Speech (TTS) models such as Coqui XTTS v2, FastSpeech, or VITS for the Somali language.

Dataset Structure

Format: .wav audio + metadata.csv (tab-separated: wav_path, transcription)
Language: Somali
Total Samples: ~20,000 clips
Sampling Rate: 22050 Hz

How to Use

To train with Coqui TTS:

tts --config_path=config.json --train_path=path/to/metadata.csv --audio_path=path/to/wavs/

Downloads last month: 1,760

Size of downloaded dataset files:

588 MB

Size of the auto-converted Parquet files:

587 MB

Number of rows:

9,884