xuducan / RapidASR

A Cross platform implementation of Wenet ASR inference. It's based on ONNXRuntime and Wenet. We provide a set of easier APIs to call wenet models.

RapidASR: a new member of RapidAI family.

Our visio is to offer an out-of-box engineering implementation for ASR.

A cpp implementation of recognize-onnx.py in Wenet-asr in which it implements the inference with ONNXRuntime. For a version of pure CPP code, we need to do a bit of work to rewrite some components.

Special thanks to its original author SlyneD.

Less is more. Less dependency, more usability.

Progress:

• Linux
• Mac
• Android
• Windows

Models

The model is original from https://github.com/wenet-e2e/wenet/tree/main/examples/wenetspeech/s0 and tested with recognize-onnx.py.

Donwload:

URL：https://pan.baidu.com/s/1BTR-uR_8WWBFpvOisNR_PA 
CODE：9xjz 

• Sample Rate:

16000Hz

• sample Depth:

16bits

• channel:

single

Build

• Linux

TBD

• Windows

TBD

Notice:

The project is under the protection of GPL V2 and commercial license.

For a commercial license, please contact us: [e-mail]

Commercial support

For a commercial user, we offer a library to resample input data including mp3, mp4, mkv and so on.

Please visit: https://github.com/RapidAI/RapidAudioKit