Multilang ASR Captioner

A multilingual automatic speech recognition and video captioning tool using faster whisper on cpu.

multilang-asr-captioner_720p.mp4

Requirements and Instalations

Docker (preferred)

You'll need to install docker.

Then, follow the steps below.

1. clone the repo
```{bash}
git clone git@github.com:marquesafonso/multilang-asr-captioner.git
```
2. Build and run the container using docker-compose

```{bash}
docker compose up
```

Check the landing page.

From there you will see the submit_video endpoint and the documentation

Tip: on Linux or Mac localhost will resolve directly to 0.0.0.0 but on windows you will need to change it to 127.0.0.1

Local

To run this tool locally on your computer you will need the following sofware installed:

• ImageMagick
• Python (3.11)

Once you are at your desired working directory, run the following commands on your terminal:

git clone git@github.com:marquesafonso/multilang-asr-captioner.git

pip install pipenv

pipenv install

Note that this assumes a proper Git installation and ssh key configuration.

Quick start (local)

API

A FastAPI API is available. This is the easiest way to use the program locally, akin to docker.

To start the API run:

pipenv run python main.py

Then check the landing page.

From there you will see the submit_video endpoint and the documentation

Tip: on Linux or Mac localhost will resolve directly to 0.0.0.0 but on windows you will need to change it to 127.0.0.1

Command Line Interface

Run the following code to use the CLI. The input file must be in mp4 format.

pipenv run python cli.py --invideo_filename '<your_file_name>' --max_words_per_line 8

Fontsize, Font, Background Color and Text Color arguments are available:

pipenv run python cli.py --invideo_filename '<your_file>' --max_words_per_line 8 --fontsize 28 --font "Arial-Bold" --bg_color None --text_color 'white'