generate_screenshot_of_video.yml

# This example shows how to fill an image field with a 'screenshot' taken from a video field of the same content.
# In doing so, it makes use of an external tool, via executing it on the command-line.
#
# It is based on the 'video' content class having 3 fields:
# - 'file': the video file itself (type: media)
# - 'screenshot': the field that will hold the screenshot (type: image)
# - 'screenshot_frame': an integer field, used to specify the time in the video at which the screenshot is taken (type: integer)

# set references to the video-specific fields
-
    type: content
    mode: load
    match:
        content_id: 1234567890
    references:
        -
            identifier: video_path
            attribute: attributes.file.path
        -
            identifier: screenshot_path
            attribute: attributes.screenshot.path
        -
            identifier: screenshot_frame
            attribute: attributes.screenshot_frame

# run 'avconv' to generate the screenshot (avconv is a fork of ffmpeg, available by default on debian 8)
-
    type: process
    mode: run
    command: avconv
    arguments:
        - '-i'
        - 'reference:video_path'
        - '-y'
        - '-ss'
        - 'reference:screenshot_frame'
        # in real-life you should add a unique token to the filename, eg. the content id. See `eval: `
        - '/tmp/test.jpg'

# upload the generated file to the screenshot field and re-publish the content
-
    type: content
    mode: update
    match:
        content_id: 1234567890
    attributes:
        screenshot:
            path: "/tmp/test.jpg"