Skip to content

A command-line tool for the ASRBench framework, simplifying audio transcription system benchmarking with a single config file, supporting popular and custom transcription systems

License

Notifications You must be signed in to change notification settings

ASRBench/asrbench-cli

Repository files navigation

🇧🇷 Versão Português Brasileiro

Asrbench-Cli

Index

Introduction

ASRBench CLI is a complementary tool to the ASRBench framework, designed to simplify the execution of audio transcription system benchmarks directly from the command line.

With it, you can:

  • Use the most popular transcription systems on the market.
  • Add new customized transcription systems.
  • Run benchmarks simply, using just one configuration file.

Installation

To install ASRBench, all you need is Python 3.12+ and pip. Use the command below to install the latest version:

pip install asrbench-cli

Note

The list of project dependencies is available in the file pyproject.toml

Usage

The CLI requires a configuration file to work, in the same format as the framework. benchmark environment, defining datasets, transcribers and output parameters in a simple and declarative way. declarative. For more details on the structure of the configuration file, go to documentation.

Below is an example of the configuration file structure:

# data output configuration
output:
  type: "csv"
  dir: "./results"
  filename: "example_filename"

# datasets configuration
datasets:
  dataset1:
    audio_dir: "resources/common_voice_05/wav"
    reference_dir: "resources/common_voice_05/txt"

# transcription system configuration
transcribers:
  faster_whisper_medium_int8:
    asr: "faster_whisper"
    model: "medium"
    compute_type: "int8"
    device: "cpu"
    beam_size: 5
    language: "en"  

With the configuration file in hand, just run the command:

asrbench-cli run path/to/configfile.yml

The CLI will read the configuration file, set up the benchmark and run it automatically. All progress and steps will be displayed directly in the terminal, including the percentage of completion and a time estimate for the completion of each stage of the transcription process.

Tip

For a complete list of available commands and instructions for more advanced uses, see documentation.

Contribution

If you want to contribute to ASRBench, see CONTRIBUTING.md for information on:

  • Configuration of the development environment.
  • Project structure and dependencies.
  • Best practices.

License

Distributed under the MIT license. See the LICENSE file for more details.

⬆️ Go to top

About

A command-line tool for the ASRBench framework, simplifying audio transcription system benchmarking with a single config file, supporting popular and custom transcription systems

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published