9min 49 sec, Quicktime
This is a section of the videos created by the headlines on Jan 10, 2007
This project is a part of the Mass Mass Media suite of projects. You can see installation views here.
In this work the headlines from Google News are translated into a continuous montage of videos from YouTube. Each word in the headline is used as keyword to search YouTube. The computer program finds the most popular video matching that word and downloads it. A section of the video is cut out depending on the length of the word and its position in the sentence. The headline is then reassembled from these video clips and played back. This project creates a visual language from the lexicon of public video sharing and the grammar of the glib language of news headlines.
Some Technical Details:
A PHP program grabs an RSS feed of the Google News main page every two minutes. These headlines are added to a database. A separate program gets the headlines from this database. The program forks (so that lots can happen at once) and, using the YouTube API, searches YouTube using each word in the headline as the search term. It downloads the most popular video, converts it from Flash to a Quicktime file and adds it to the video cache. Now here is where it gets a bit sticky, the program then grabs one second of video for each letter in the word, starting at the word's position in the headline. It does this for each word in the headline. It then splices all of the videos back together and uses Applescript to add it to an iTunes playlist. iTunes is then told to start playing (if it was not already).
Example:
Headline: "Europe Relieved as Crisis Is Averted"
For the word "Relieved" the program would search YouTube for "Relieved" and download the "Most Popular" (according to YouTube) video. That gets converted to quicktime. The total play length of the downloaded video is divided into 6 sections (since there are 6 words in the headline). It then grabs 8 seconds of video from the second section (8 letters in "Relieved" and it is the second word). the program does this for each word in each headline. The processes are forked so that it can download and process multiple videos at once and really take advantage of a Core Duo Processor.
Source code is available here
Many thanks to:
youtube-dl.sh python script for downloading YouTube Videos.
catmovie which is part of QTCoffee a helpful suite of command line programs for working with quicktime files.
ffmpeg a great mpeg tool.