AIbase
EN
Home
Wav2vec2 Xls R 1b Hebrew
W

Wav2vec2 Xls R 1b Hebrew

Developed by imvladikon
An automatic speech recognition model fine-tuned on Hebrew datasets based on facebook/wav2vec2-xls-r-1b
Speech Recognition
Transformers
 OtherOpen Source License:Apache-2.0#Hebrew speech recognition#Large model fine-tuning#Multi-GPU training
Downloads 66
Release Time : 3/2/2022

Model Overview

This model is an automatic speech recognition (ASR) model optimized for Hebrew, fine-tuned from Facebook's wav2vec2-xls-r-1b architecture, suitable for Hebrew speech-to-text tasks.

Model Features

Hebrew optimization
Specially fine-tuned and optimized for Hebrew speech recognition tasks
Large-scale pretraining foundation
Fine-tuned from Facebook's 1-billion-parameter wav2vec2-xls-r model
Efficient training
Utilizes multi-GPU distributed training and mixed-precision training techniques

Model Capabilities

Hebrew speech recognition
Robust speech event detection
Speech-to-text

Use Cases

Speech transcription
Hebrew meeting minutes
Automatically transcribe Hebrew meeting recordings into text records
Word Error Rate (WER) 22.51%
Hebrew voice assistant
Provide speech recognition capabilities for Hebrew voice assistants
Education
Hebrew learning applications
Used for pronunciation evaluation features in Hebrew learning applications
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