zoom_in Click to enlarge

WeNet

▲ ▼

language

description WeNet Overview

WeNet is an open source speech-to-text software toolkit designed for advanced Automatic Speech Recognition (ASR) research and production. Developed by Tencent, it utilizes neural networks to process streaming audio. The modular design supports various architectures making it suitable for researchers, developers, and organizations needing customizable ASR solutions.

help WeNet FAQ

Can one WeNet model handle both live streaming and offline transcription?

Yes. WeNet's Unified Two-pass architecture was designed to support streaming and non-streaming decoding within one model.

Is WeNet a ready-made dictation app like Otter or Dragon?

No. It is an open-source ASR toolkit and runtime for researchers and developers who want to train, adapt, or deploy speech-recognition systems.

Which languages and pretrained models can WeNet run?

The project provides or connects to models for Chinese and English, including Paraformer and Whisper-based choices in its current Python interface. Language coverage depends on the selected pretrained model or the data used for training.

Can WeNet be deployed without a Python server in production?

Yes. The project includes a lightweight runtime intended for deployment, in addition to its Python training and inference tools.

swap_horiz

Looking for WeNet alternatives? Compare top competitors ranked & scored

arrow_forward