W

Wav2vec2 Xls R 1b En To 15

Developed by facebook
Facebook's Wav2Vec2 XLS-R model fine-tuned for speech translation tasks, supporting translation from English to 15 target languages.
Downloads 505
Release Time : 3/2/2022

Model Overview

This model is a speech encoder-decoder model capable of translating spoken English into 15 different written languages. The encoder is based on facebook/wav2vec2-xls-r-1b, the decoder on facebook/mbart-large-50, and it has been fine-tuned on the Covost2 dataset.

Model Features

Multilingual support
Supports speech translation from English to 15 different languages.
XLS-R architecture
Utilizes the XLS-R architecture with large-scale self-supervised learning to provide high-quality speech representations.
End-to-end translation
Directly generates target language text output from speech input without intermediate transcription steps.

Model Capabilities

English speech recognition
Multilingual text generation
Speech-to-text translation

Use Cases

Speech translation
Real-time speech translation
Translates spoken English into multiple target languages in real-time.
Performs excellently on the Covost2 dataset.
Multilingual subtitle generation
Automatically generates multilingual subtitles for English video content.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase