autocrop
Automatically detects and crops faces from batches of pictures or video frames.
Open-source Python tool and CLI for automated, face-centered image and video frame cropping.
Autocrop is a Python library and command-line utility that detects the biggest face in an image and crops around it using OpenCV's YuNet neural network. It suits video editors and media producers handling batch portrait processing or video frame extraction pipelines.
What it is
The tool ingests image files or standard input, identifies faces using a bundled ONNX model, and outputs resized or native-resolution crops centered on the subject. It integrates cleanly with shell utilities like find, fd, and GNU parallel for batch workflows, or directly via a Python API.
Who it's for
It suits video editors and developers integrating automated face framing into command-line scripts or Python pipelines. It does not suit users needing a graphical user interface or interactive face-selection tools.
Key features
OpenCV YuNet Detection
Uses a bundled neural network face detector that operates completely offline.
Flexible CLI and API
Accepts single images, standard input, or NumPy arrays directly through Python.
Video Frame Processing
Extracts video frames via FFmpeg and crops faces across sequences using shell loops.
Custom Dimensions
Allows explicit width, height, and face percentage scaling parameters.
Up and running in 3 steps
Our honest verdict
What's great
- Bundles the YuNet ONNX model locally so it runs entirely offline without runtime model downloads
- Works seamlessly with standard stream piping and command-line batch tools like find and GNU parallel
- Accepts both file paths and OpenCV-style NumPy arrays via a straightforward Python API
- Supports dozens of image formats including JPEG, PNG, WebP, and TIFF
Where it falls short
- Processes strictly one image per standard CLI invocation, requiring custom shell scripts for batch runs
- Only handles the first frame of animated GIFs and ignores subsequent frames
- Requires familiarity with command-line scripting and FFmpeg to process video files effectively
- The repository metadata does not explicitly state an open-source license for the package itself
Pricing & license
The software is available as an open-source package with no hosted service options mentioned. The only cost is your own hardware and execution time.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
How does autocrop handle video files?
Autocrop is designed for still images, but you can extract video frames using FFmpeg and pipe them through autocrop in a shell loop.
Can I run autocrop without an internet connection?
Yes. It vendors OpenCV Zoo's YuNet face detection model directly, allowing entirely offline processing.
What inputs does the Python API accept?
The Cropper crop method accepts standard filepaths or OpenCV-style BGR and BGRA NumPy arrays.
