Installation

In this section, we describe how to install sherpa-onnx on various platforms.

Requirements:

• CMake >= 3.13

• A compiler that supports at least C++14

Hint

You can use pip install cmake to install the latest cmake.

• Linux
  • CPU (Linux x64 or Linux arm64)
  • GPU (CUDA 11.8, CUDNN 8, Linux x64)
  • GPU (CUDA 12.8, CUDNN 9, Linux x64)
  • GPU (CUDA 10.2, CUDNN8, Linux arm64, e.g., Jetson Nano B01)
  • GPU (CUDA 11.4, CUDNN8, Linux arm64, e.g., Jetson Orin NX)
  • GPU (CUDA 12.6, CUDNN9, Linux arm64, e.g., Jetson Orin Nano Engineering Reference Developer Kit Super Jetpack 6.2)
• macOS
• Windows
  • Download pre-compiled libs and binaries for Windows
    • Pre-compiled executables and libraries for Windows x64
      • Shared Libraries
      • Static Libraries
    • Pre-compiled executables and libraries for Windows x86
      • Shared Libraries
      • Static Libraries
    • Pre-compiled executables and libraries for Windows arm64
      • Shared Libraries
      • Static Libraries
  • Build from source for Windows (CPU only)
    • Build sherpa-onnx on Windows (x64)
      • Prerequisites
      • Build Configurations
        • Shared + MD + Release
        • Shared + MD + Debug
        • Shared + MD + MinSizeRel
        • Shared + MD + RelWithDebInfo
        • Shared + MT + Release
        • Shared + MT + Debug
        • Shared + MT + MinSizeRel
        • Shared + MT + RelWithDebInfo
        • Static + MD + Release
        • Static + MD + Debug
        • Static + MD + MinSizeRel
        • Static + MD + RelWithDebInfo
        • Static + MT + Release
        • Static + MT + Debug
        • Static + MT + MinSizeRel
        • Static + MT + RelWithDebInfo
    • Build sherpa-onnx on Windows (x86)
      • Prerequisites
      • Build Configurations
        • Shared + MD + Release
        • Shared + MD + Debug
        • Shared + MD + MinSizeRel
        • Shared + MD + RelWithDebInfo
        • Shared + MT + Release
        • Shared + MT + Debug
        • Shared + MT + MinSizeRel
        • Shared + MT + RelWithDebInfo
        • Static + MD + Release
        • Static + MD + Debug
        • Static + MD + MinSizeRel
        • Static + MD + RelWithDebInfo
        • Static + MT + Release
        • Static + MT + Debug
        • Static + MT + MinSizeRel
        • Static + MT + RelWithDebInfo
    • Build sherpa-onnx on Windows (ARM64)
      • Prerequisites
      • Build Configurations
        • Shared + MD + Release
        • Shared + MD + Debug
        • Shared + MD + MinSizeRel
        • Shared + MD + RelWithDebInfo
        • Shared + MT + Release
        • Shared + MT + Debug
        • Shared + MT + MinSizeRel
        • Shared + MT + RelWithDebInfo
        • Static + MD + Release
        • Static + MD + Debug
        • Static + MD + MinSizeRel
        • Static + MD + RelWithDebInfo
        • Static + MT + Release
        • Static + MT + Debug
        • Static + MT + MinSizeRel
        • Static + MT + RelWithDebInfo
  • Build from source for Windows (With NVIDIA GPU)
    • GPU (CUDA 11.8)
    • GPU (CUDA 12.x)
• Embedded Linux (aarch64)
  • Install toolchain
  • Build sherpa-onnx
  • sherpa-onnx-alsa
  • How to build static libraries and static linked binaries
• Embedded Linux (arm)
  • Install toolchain
  • Build sherpa-onnx
  • How to build static libraries and static linked binaries
• Embedded Linux (riscv64)
  • Install toolchain
  • Build sherpa-onnx
  • qemu
    • Run speech-to-text with qemu
    • Run text-to-speech with qemu

If you want to build an Android app, see Android. If you want to build an iOS app, see iOS.