---
license: cc0-1.0
language:
- lv
pipeline_tag: text-to-speech
---
# Latvian TTS dataset "Rūdolfs"

Dataset is created from [audio books](https://www.youtube.com/@LatvijasNeredzigobiblioteka) of 
[Latvijas Neredzīgo bibliotēka](https://neredzigobiblioteka.lv/). For permission reasons the original voice
in the recordings has been cloned to a donor voice. 
Audio book recordings are used with permission from Latvijas Neredzīgo bibliotēka.

## Optional processing

For some tastes the "s" sounds in the clips released may feel too harsh. 
To reduce them use the `sibilant_reducer.py` script.

```commandline
uv run python sibilant_reducer.py --input-dir ../clips/wav --output-dir ../clips/wav_deessed
```