Whisper-large-v2-hi-v3 Open-source Hindi Speech Recognition Model - Accurately Recognize Speech with Low Error Rate

Whisper Large V2 Hi V3

Developed by anuragshas

Hindi speech recognition model fine-tuned based on OpenAI Whisper Large-v2, achieving a word error rate of 11.3% on the Common Voice 11.0 Hindi test set

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Hindi Speech Recognition #Low Word Error Rate #Multi-GPU Training

Downloads 21

Release Time : 12/15/2022

Model Overview

This is an automatic speech recognition (ASR) model specifically optimized for Hindi, based on OpenAI's Whisper architecture and fine-tuned on the Mozilla Common Voice dataset

Model Features

Low Word Error Rate

Achieves an excellent word error rate of 11.3% on the Hindi test set

Multi-GPU Training Optimization

Supports distributed multi-GPU training with gradient accumulation techniques to improve training efficiency

Fine-tuning

Performed 1000 steps of fine-tuning on the Common Voice Hindi dataset

Model Capabilities

Hindi Speech Recognition

Speech-to-Text

Long Audio Processing

Use Cases

Speech Transcription

Hindi Speech Transcription

Convert Hindi speech content into text

11.3% word error rate

Voice Assistants

Hindi Voice Command Recognition

Provides speech recognition capabilities for Hindi voice assistants

Training Loss	Epoch	Step	Validation Loss	Wer
0.0479	2.06	200	0.2189	12.3226
0.0081	5.06	400	0.2649	11.5740
0.001	8.06	600	0.2998	11.4252
0.0004	11.05	800	0.3191	11.3039
0.0003	14.05	1000	0.3267	11.3291

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご

Whisper Large V2 Hi V3

Model Overview

Model Features

Model Capabilities

Use Cases

🚀 Whisper Large-v2 Hindi

🚀 Quick Start

📚 Documentation

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

📄 License