WeNet

WeNet is a speech recognition toolkit for developers adding transcription to applications or services. It covers model training and production deployment, so teams can adapt a recognizer to their audio or start with a pretrained model.

• Model training: Train Transformer-based speech recognition models with workflows for LibriSpeech, AIShell and large datasets.
• Pretrained recognition: Use existing speech models with a production runtime when you don't need to train your own.
• Specialized vocabulary: Context biasing and language model integration help recognition account for terms relevant to your application.
• Application integration: Python and command-line interfaces let developers incorporate recognition into their software, with JIT support for deployment.

WeNet uses PyTorch and runs on your own hardware, with CPU, CUDA GPU and Ascend NPU support. It's open source under the Apache 2.0 license.