screenshotgenerator 0.3.0

A library for automatically generating and selecting video screenshots.

screenshotgenerator

This project aims to automate the selection of video screenshots. It employs ffmpeg to generate a pool of screenshots then calls on two autogluon machine learning models to score the screenshots in order to select the best of them. The first model attempts to determine whether a screenshot is focused while the second model attempts to determine whether the screenshot is a portrait (a close-up of one or more people). The focused model's score makes up 75% of the total score while the portrait model's score makes up 25% of the total score.

Dependencies

The following must be installed on your system:

• ffmpeg
• MediaInfo
• Microsoft Visual C++ Redistributable (if you're on Windows)

CUDA

autogluon uses the CPU version of PyTorch, by default. If you have a CUDA-enabled GPU, installing the CUDA version of PyTorch may increase prediction speed:

pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116

Usage

Download and extract models.zip.

CLI

Usage: screenshotgenerator [OPTIONS]

Options:
  --end-time [%H:%M:%S]           The time at which to stop taking
                                  screenshots. Defaults to 95% of the video
                                  duration, to exclude credits.
  --ffmpeg-path TEXT              The path to ffmpeg. Defaults to 'ffmpeg', 
                                  which requires ffmpeg to be in your path.
  --models-directory TEXT         The path to the directory containing the
                                  autogluon models. If the directory doesn't
                                  exist, the pretrained models will be
                                  downloaded to this location. Defaults to a 
                                  subdirectory in the user data directory.
  --pool-directory TEXT           The directory in which to store the
                                  screenshot pool. Defaults to the temporary 
                                  directory.
  --pool-report-path TEXT         A JSON file detailing the screenshot pool,
                                  sorted by descending preference.
  --pool-size INTEGER             The size of the pool from which to select
                                  screenshots.  [default: 64]
  --portrait-preference [mixed|noportrait|portrait]
                                  Preference regarding portrait screenshots.
                                  [default: portrait]
  --screenshot-count INTEGER      The number of screenshots to select.
                                  [default: 4]
  --screenshot-directory TEXT     The directory into which to copy the
                                  selected screenshots.  [required]
  --silent                        Suppress ffmpeg and autogluon output.
  --start-time [%H:%M:%S]         The time at which to start taking
                                  screenshots.  [default: 00:00:00]
  --video-path TEXT               The path to the video for which to generate
                                  screenshots.  [required]
  --help                          Show this message and exit.

Example

screenshotgenerator --screenshot-directory "B:\Screenshots" --video-path "Z:\Encodes\A Great Movie (2023).mkv --portrait-preference mixed

Library

The library provides the following:

• generate function
• PortraitPreference enum
• Screenshot class

Parameters

• screenshot_directory: str
  The directory into which to copy the selected screenshots.
• video_path: str
  The path to the video for which to generate screenshots.
• end_time: datetime
  The time at which to stop taking screenshots. Defaults to 95% of the video duration, to exclude credits.
• ffmpeg_path: str
  The path to ffmpeg. Defaults to 'ffmpeg', which requires ffmpeg to be in your path.
• models_directory: str The path to the directory containing the autogluon models. Defaults to a subdirectory in the user data directory.
  If the directory doesn't exist, the pretrained models will be downloaded to this location.
• pool_directory: str
  The directory in which to store the screenshot pool. Defaults to the temporary directory.
• pool_size: int
  The size of the pool from which to select screenshots. Defaults to 64.
• portrait_preference: PortraitPreference
  Preference regarding portrait screenshots. Defaults to PortraitPreference.PORTRAIT.
• screenshot_count: int
  The number of screenshots to select. Defaults to 4.
• silent: bool
  True to suppress ffmpeg and autogluon output. Defaults to false.
• start_time: datetime
  The time at which to start taking screenshots. Defaults to 00:00:00.

Returns

list[Screenshot] The screenshot pool, sorted by descending preference.

Example

import screenshotgenerator

screenshots = screenshotgenerator.generate(
    screenshot_directory=r"B:\Screenshots",
    video_path=r"Z:\Encodes\A Great Movie (2023).mkv",
    portrait_preference = screenshotgenerator.PortraitPreference.MIXED)