ocrvid 0.5.0

CLI tool for OCR-ing video frames on macOS

ocrvid

CLI tool to extract text from videos using OCR on macOS.

[!NOTE] Currently, this tool only tested and works on macOS 13 or later.

[!CAUTION] This tool is still in early development stage. Current v0.x releases are not stable and may have breaking changes.

Installation

Install this tool using pip:

pip install ocrvid

Usage

Run OCR on a video

To extract text from a video, run:

ocrvid run path/to/video.mp4

then ocrvid generates frames from the video and runs OCR on each frame. Optionaly, frames can be saved in a directory passed --frames_dir / -fd.

OCR results are saved in a json file named video.json in the current directory. (where video is taken from input file name video)

for example, run against the test video file at tests/video/pexels-eva-elijas.mp4 in this repo:

ocrvid run tests/video/pexels-eva-elijas.mp4

Then pexels-eva-elija.json is generated in the current directory which looks like this:

{
    "video_file":"tests/video/pexels-eva-elijas.mp4",
    "frames_dir":"xxx/.ocrvid/frames/pexels-eva-elijas",
    "frame_step":100,
    "frames":[
        {
            "frame_file":"xxx/.ocrvid/frames/pexels-eva-elijas/frame-0.png",
            "results":[
                {
                    "text":"INSPIRING WORDS",
                    "confidence":1.0,
                    "bbox":[
                        0.17844826551211515,
                        0.7961793736859821,
                        0.3419540405273438,
                        0.10085802570754931
                    ]
                },
                {
                    "text":"\"Foar kills more dre",
                    "confidence":1.0,
                    "bbox":[
                        0.0724226723609706,
                        0.6839455987759758,
                        0.4780927975972494,
                        0.14592710683043575
                    ]
                },
                {
                    "text":"than failure ever",
                    "confidence":1.0,
                    "bbox":[
                        0.018455287246445035,
                        0.6549868414269003,
                        0.45329265594482426,
                        0.14363905857426462
                    ]
                },
                {
                    "text":"IZY KASSEM",
                    "confidence":0.5,
                    "bbox":[
                        -0.015967150208537523,
                        0.6675747977206025,
                        0.23065692583719888,
                        0.08114868486431293
                    ]
                },
                {
                    "text":"Entrepreneur",
                    "confidence":1.0,
                    "bbox":[
                        0.01941176222542875,
                        0.1353812367971159,
                        0.9058370590209961,
                        0.26137274083956863
                    ]
                }
            ]
        },
...

Interact with YouTube

Interacting YouTube? Please see yt-dlp.

Development

To contribute to this tool, first checkout the code. Then create a new virtual environment:

cd ocrvid
python -m venv venv
source venv/bin/activate

Now install the dependencies and test dependencies:

pip install -e '.[test,dev]'

To run the tests:

make test