AIbase
EN
Home
Whisper Small Ko Zeroth
W

Whisper Small Ko Zeroth

Developed by seastar105
A Korean automatic speech recognition model fine-tuned on the Zeroth Korean dataset, with a word error rate of 6.76%.
Speech Recognition
Transformers
 KoreanOpen Source License:Apache-2.0#Korean speech recognition#Low word error rate#Fine-tuning Whisper
Downloads 285
Release Time : 12/11/2022

Model Overview

This model is an optimized version of OpenAI Whisper-small for Korean, specifically designed for high-accuracy Korean speech-to-text tasks.

Model Features

Korean optimization
Specifically fine-tuned for the characteristics of Korean speech to improve recognition accuracy.
Low word error rate
Achieved a word error rate of 6.76% on the evaluation set, showing excellent performance.
Efficient training
Using mixed-precision training, good results can be achieved in 4000 steps.

Model Capabilities

Korean speech recognition
Speech-to-text
Audio content transcription

Use Cases

Speech transcription
Korean meeting records
Automatically convert Korean meeting recordings into text records.
Accuracy of 93.24%
Media subtitle generation
Automatically generate subtitles for Korean video content.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase