---
license: cc-by-4.0
language:
- tr
task_categories:
- text-to-speech
- audio-to-audio
tags:
- speech
- codec
size_categories:
- 100K<n<1M
---

# Turkish codec training packs

About 334 hours of Turkish speech, re-packed for training a speech codec. Audio is 24 kHz mono FLAC, stored as
`alania-NNNNN.pack` (concatenated FLAC clips) with `alania-NNNNN.idx.npy` (offset, byte length, sample count per clip).

## Source and licence

All clips come from the `cc-by` subset of
[cloud0day3/alania-synthetic-speech-tr](https://huggingface.co/datasets/cloud0day3/alania-synthetic-speech-tr)
(revision `30639faf6a603af982dec98b0ea47e22cb7c7946`), licensed CC-BY 4.0. Clips were resampled to 24 kHz and
re-encoded; nothing else was changed. Please credit the original dataset when you use this one.