OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Fastasr

Fastasr is an Apache-2.0-licensed C++ library for automatic speech recognition inference. It implements three non-streaming models and one streaming model, and provides a static library for C++ applications plus a Python module.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Fastasr — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Fastasr when you need an embeddable inference library and can select between sentence-based recognition and streaming results. Its documented ARM example makes it relevant for developers evaluating speech recognition on Raspberry Pi 4B.

Who it is for

  • C++ developers adding speech-to-text inference to an application.
  • Python developers using the documented PyFastASR module.
  • Developers considering local inference on ARM hardware such as Raspberry Pi 4B.

Documented uses

  • Transcribe speech in sentence-based batches with a non-streaming model.
  • Process incoming speech and return recognition results through the streaming model.
  • Handle longer audio with non-streaming models using VAD.
Before you choose

Requirements & limits

The public repository specifies an Apache-2.0 license. The README documents a Python package and source-build paths; no hosted service is described.

  • The streaming conformer_online model returns results from speech input in real time; the README notes an accuracy trade-off against non-streaming models.
  • The three non-streaming models operate sentence by sentence and are described as less immediate but more accurate; VAD is documented for longer audio.
  • The listed model languages are Chinese and English for Paraformer, and Chinese for k2_rnnt2, conformer and conformer_online.
  • The README lists libfftw3 and libopenblas as third-party dependencies and says inference does not depend on PaddlePaddle, PyTorch or TensorFlow.
  • The README documents Python package support for CPython 3.6–3.11 on listed platforms, with source compilation for other platforms.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Build FastASR from source on Linux

Compile the FastASR C++ project on Linux.

Before you start

  • Linux environment
  • libfftw3 and libopenblas development libraries
  • Python 3 and Python development files

Steps

  1. Install the dependencies: sudo apt-get install libfftw3-dev libfftw3-single3.

  2. Install OpenBLAS and Python development packages: sudo apt-get install libopenblas-dev and sudo apt-get install python3 python3-dev.

  3. Clone the repository with the command below.

    Documented command
    git clone https://github.com/chenkui164/FastASR.git

  4. Build the C++ project: cd FastASR/, mkdir build, cd build, cmake -DCMAKE_BUILD_TYPE=Release .., then make.

Expected outcome

The README’s Linux build procedure produces a compiled FastASR project; it also describes a separate command for building a Python wheel.

Guide limitations

  • This procedure covers building, not running inference or integrating a model.
  • The README does not specify additional Linux distributions or build troubleshooting.
Documentation 1 ↗
Source-based practical guide

Build a FastASR Python wheel from source on Linux

Create a Python wheel after obtaining the FastASR source and installing its build dependencies.

Before you start

  • Linux environment
  • FastASR source cloned as described in the Linux build guide
  • Python 3 and Python development files, plus libfftw3 and libopenblas development libraries

Steps

  1. From the repository root, run the command below.

    Documented command
    python -m build

  2. Use the built wheel as the package artifact; the README separately documents direct installation with the command below.

    Documented command
    pip install fastasr

Expected outcome

A Python wheel is built from the source tree.

Guide limitations

  • The README does not provide wheel installation steps for the locally built artifact or Python usage examples in the documented section.
  • This procedure does not cover inference execution.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
chenkui164/FastASR · Official homepage title · retained in its source language
How we review sourcesSuggest a correction