cvml-sp21

FFMPEG Guide

Videos

Extracting images from a video file (and resizing/cropping them to 1024x1024).

After following this tutorial, your images will be ready to be zipped and used to train a StyleGAN2 model.

Installing FFMPEG

Using FFMPEG

  1. Open Terminal in Mac, or Command Prompt in Windows
  2. cd into the directory with your input media
  3. run the desired FFMPEG command

Common FFMPEG commands

Note: The following commands can be modified to suit your needs. I’ll put some common “flags” here that you’ll see being used below:

  1. -r: the framerate (e.g.: -r 30 for 30fps)
  2. -i: the input file(s). Just something you always need.
  3. -crf: the “quality” when exporting to .mp4. Higher number = lower quality (more compression, smaller file sizes). Stick between 17-25.

Note: If you’re extracting images from a video, you can save them as PNG or JPG.

  1. Turn an image sequence into an .mp4 (60 fps, scale is the same as input images)

    ffmpeg -r 60 -i %d.jpg -vcodec libx264 -crf 23 -pix_fmt yuv420p output.mp4

  2. Extract every frame from a video

    ffmpeg -i input.mp4 -vf frames/output%d.jpg

  3. Extract a certain number of frames from a video by setting fps

    ffmpeg -i input.mp4 -vf fps=1 frames/output%d.jpg

  4. Extract frames from video, also apply resize, crop, and fps. ffmpeg -i input.mp4 -q:v 2 -vf scale=-1:1024,fps=1,"crop=1024:1024" frames/output%5d.jpg

  5. Same as above, but with a start offset so nothing from the first 30 seconds of video is captured: ffmpeg -i input.mp4 -ss 30 -q:v 2 -vf scale=-1:1024,fps=1,"crop=1024:1024" frames/output%5d.jpg

More Information:

Cutting small sections FFMPEG Upscaling Tutorial