
600 iterations, take the last frame of that effort, and create a cross-fade to the original full-res poster. I could let primitive run for thousands of frames, it would take forever, but the final image would never be a 100% match with the original poster. I use the font Amatic SC Bold from Google Fonts,Īnd I let the text fade out after 3 seconds.Īt the end of the video, I do want to show the unmodified full resolution movie poster, so that the solution is always clear.

filter_complex "split drawtext=text=' $opening ':įontcolor=black:fontsize=120:fontfile=fonts/AmaticSC-Bold.ttf:x=(w-text_w)/2:y=(h-text_h)/2,format=yuv420p,fade=t=out:st=3:d=1:alpha=1 overlay" -profile:v main -level 3.1 -preset medium -crf 23 -x264-params ref =4 -movflags +faststart -y " $output " I then look up the IMDB page for this movie, use again htmlq to get the movie title and poster.įfmpeg -i " $input " -vcodec libx264 -pix_fmt yuv420p -ss 1 -t 40 -r 12 I use movie list URLs like /chart/boxoffice as a source, parse the HTML with the Rust tool htmlq and pick one movie at random. What better place to get movie posters from than IMDB? I wanted to be able to download movie posters of specific movies or let my script select one at random. In the end, I automated the whole process in a bash script, based on myīashew bash framework: /pforret/squaredforwork Step 1: get a photo

This time I wanted to take it a bit further and make a video where the image becomes clearer by the secondĪnd there is a challenge in guessing as fast as possible what movie is being shown. Some examples of earlier tests might give you an idea (working with large square pixels here: I mostly worked with movie posters, since those are often universally recognisable images. It is based on years of experimenting with image manipulation and information reduction.īasically: how much visual information do you need to recognise a (familiar) image? One of the projects I started during the lockdown is squaredforwork Creating image reveal videos with ffmpeg and primitive (bash)
