vosk-cli

This python package serves as an Vosk interface for Opencast. It allows to generate subtitles (WebVTT files) from Video and Audio sources via Vosk.

Installation

1. Install vosk-cli

To install the latest stable version of vosk-cli, run

pip install vosk-cli

Alternatively, to install the latest development version, clone this project and inside the project directory run

pip install .

2. Install dependencies

• FFmpeg
• ffprobe

Vosk-cli uses ffprobe to analyze and ffmpeg to preprocess input files. The easiest way to install ffmpeg is by using a package manager. If you want or need to install from source, visit FFmpeg.org/download.html and follow the instructions for your operating system.

3. Download the language model

Go to https://alphacephei.com/vosk/models and download at least the English language model. The larger models generally yield better results.

Usage

Now you are able to run vosk-cli -i <input_file_path> -o <output_file_path> -m <model_name_or_path>.

For example, if there is a video.mp4 file in your download folder and a model named vosk-model-en-us-0.22 in the ./models folder you created, you can run

vosk-cli -i ~/Downloads/video.mp4 -o text -m vosk-model-en-us-0.22

This will create a text.vtt file (which contains the transcribed captions) in your current directory.

Model Search Paths

Vosk-cli follows the following algorithm when attempting to find matching models:

• Check if the -m flag is passed as an absolute path, if so use it verbatim.
• Check if ./models/$model is a directory, if so use it.
• Check if /usr/share/vosk/models/$model is a directory, if so use it.

NOTE: Vosk-cli requires an exact match between directory and model name. If you pass -m eng you must have an eng model in your model search path!