Streaming ASR

This page describes how to use sherpa for streaming ASR.

The following types of transducer models are supported:

• Emformer

• ConvEmformer

• LSTM

• Zipformer

We support standalone speech recognition as well as server/client based speech recognition using WebSocket.

• Standalone
  • Transducer
    • Download the pre-trained model
    • greedy search
    • modified beam search
    • fast_beam_search
    • fast_beam_search with LG
• Endpointing
  • Endpointing Demo (English)
  • Endpointing Demo (Chinese)
    • Short Demo
    • Long Demo
  • Endpointing Demo (Arabic)
• Secure connections
  • Generate a certificate
  • Start the https server
  • Start the websocket server
• Emformer transducer based streaming ASR
  • Server
    • Usage
  • Client
    • Usage
• ConvEmformer transducer based streaming ASR
  • Server
    • Usage
  • Client
    • Usage
• Conformer transducer based streaming ASR
  • Streaming Chinese ASR
    • Server
    • Client
  • Streaming English ASR
    • Server
    • Client
• LSTM transducer based streaming ASR
  • English
    • Server
    • Client
  • Chinese
    • Server
    • Client