Skip to content

Dataset · KimTaeHyun1

Multi-hour Korean speech corpus and annotations for TTS

KimTaeHyun1/MAGO_share

Large collection of Korean audio recordings (thousands of hours) with metadata, tags, and instructions for speech synthesis.

What's inside

The dataset combines three verified Korean data packages: AIHub_2k (over 1.66 million records, about 2,042 hours of audio in WAV/FLAC format with JSON instruction and tag files), Gemini_4p7k (over 1 million records, about 4,721 hours), and Gemini_5k_KO. Materials are packaged in TAR archives, with Hugging Face table splits serving as a navigation index.

What you can do with it

  • Training and fine-tuning Korean text-to-speech (TTS) models
  • Preparing automatic speech recognition (ASR) systems for Korean on large volumes of data
Tasks
Text → speech
Languages
Korean
Size
1M–10M rows
Downloads this month
28.2K
Likes
0
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("KimTaeHyun1/MAGO_share")