Flamme-VRM/kazakh-speech-dataset

Kazakh Speech Dataset

If you find this dataset helpful please press 'like' button

Summary

The Kazakh Speech Dataset is a large-scale open-source speech corpus for the Kazakh language. This dataset is designed to support the development of Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) systems for the Kazakh language.

Dataset Statistics

• —Total Audio Duration: ~726 hours
• —Language: Kazakh (kk)
• —Audio Format: FLAC
• —Sampling Rate: 16kHz

Dataset Structure

Origin: KSC2

The dataset consists of crowdsourced audio recordings and their corresponding text transcriptions.

• —Audio: Located in crowdsourced/
• —Metadata: The train.csv file contains metadata for each recording, including:
• —audio_path: Path to the audio file
• —text: Transcription of the audio
• —duration: Duration of the audio in seconds
• —sample_rate: Sampling rate of the audio

Usage

This dataset is compatible with the Hugging Face datasets library.

from datasets import load_dataset

# Загрузка датасета
dataset = load_dataset("Flamme-VRM/kazakh-speech-dataset")

# Пример доступа к данным
print(dataset["train"][0])

Original License: CC BY 4.0 Modifications: VAD-based audio slicing, quality filtering, re-transcription with Whisper Turbo(Fine-Tuned)

Citation

@misc{flamme_vrm_kazakh_speech_2026,
  author    = {Flamme-VRM},
  title     = {Kazakh Speech Dataset (Optimized ASR Version)},
  year      = {2026},
  publisher = {Hugging Face},
  note      = {Extensive re-processing with VAD, quality filtering, and Whisper Turbo re-transcription},
  howpublished = {\url{https://huggingface.co/datasets/Flamme-VRM/kazakh-speech-dataset}}
}