Skip to main content

Martha Graham's Google doodle is not a video, is an image (a CSS sprite)


Today's Google doodle is featuring Martha Graham, an American modern dancer. The new doodle looks like a movie, but in fact it is only a CSS sprite.
The sprite is this one: 
You may be wondering how is the animation/movie appearance achieved if it only uses this static image? 

The technique involves CSS sprites. A sprite is a composition of several images grouped together in order to reduce the number of required HTTP request to the server. This way a website with 50 images requires only one request for the large image composition. The entire doodle involves only one sprite, which speeds up loading a lot (compared to loading each frame separately). Google has used this technique to leverage data transfer from the servers and consequently make the animation load faster than a traditional video clip. 

So how is this sprite converted to an animation?

Pretend you are looking through pierced piece of paper, so that you can only look through the hole in the paper. Right behind the paper there is a film strip. If you move the film strip quickly from one side to the other you will think that there is actual motion happening. This is basically how movies are made.
In CSS all you have to do is set the width and height of the container (the hole in the paper) and change the background-position at every frame. Google has done it using many div tags. 
The first div tag looks like this:
<div>&nbsp;</div> <pre><div id="hplogo0" style="left: 307px; top: 48px; width: 88px; height: 89px; background-image: url(; background-attachment: initial; background-origin: initial; background-clip: initial; background-color: initial;  background-position: 0px 0px; background-repeat: no-repeat no-repeat;"> </div>
The second:
<div id="hplogo1" style="left: 307px; top: 48px; width: 89px; height: 89px; background-image: url(; background-attachment: initial; background-origin: initial; background-clip: initial; background-color: initial; background-position: -88px 0px; background-repeat: no-repeat no-repeat;"> </div>

Notice how the background position has been shifted, so that the next frame should look like this: 
Changing the background position of the sprite very quickly therefore creates the illusion of animation.

Because Google tries hard to optimize the sizes of these doodles (that's why they did not simply create a video file) they have not even maintained equal ratios between the sample frames which would make it very simple to write a javascript function to move the background at each new frame (every 83 milliseconds). 

Therefore they are storing an array of coordinates where the next images should be positioned. I have copied their code and added a few comment to make it a bit more readable.
//This is the array of coordinates of each individual frame var d=[[307,48,88,89],[307,48,89,89],[307,48,91,89],[305,49,93,89],[305,50,93,88],[305,50,93,88],[306,52,92,86],[305,53,93,84],[305,54,94,83],[306,54,93,83],[307,54,92,83],[307,54,92,83],[308,54,90,83],[308,54,90,83],[306,53,91,84],[306,53,91,84],[308,53,90,84],[308,53,90,84],[305,53,92,84],[305,52,92,85],[306,52,91,85],[308,51,88,87,1],[308,50,88,88],[308,49,88,88],[307,49,89,88],[307,50,89,87],[308,51,89,86],[307,54,90,83],[307,57,90,80],[306,58,92,79],[306,58,92,79],[305,60,92,77],[302,61,95,76],[302,63,95,74],[302,51,96,86],[302,66,98,71],[304,67,96,69],[301,63,96,74],[301,58,93,79],[291,52,94,85],[288,50,71,88],[285,43,76,95],[285,37,70,101],[281,29,55,109],[278,20,58,119],[278,20,55,119,1],[277,12,121,127],[271,2,122,138],[267,1,126,139],[264,0,136,140],[260,0,141,140],[255,0,148,140],[252,0,151,140],[249,2,121,138],[247,3,123,137],[246,3,123,137],[246,2,124,137],[258,2,112,137],[263,2,106,137],[263,2,106,137],[262,2,103,137],[260,2,104,136],[260,2,104,137,1],[268,2,98,137],[267,2,99,137],[266,2,97,137],[266,3,96,136],[264,3,99,136],[263,3,100,136],[261,3,100,136],[259,2,138,137],[254,2,126,137],[247,2,101,136],[240,2,108,136],[238,1,110,137],[230,1,118,138],[220,15,128,124],[211,18,137,121],[205,43,102,96],[202,45,104,93],[200,38,97,101],[198,38,104,101,1],[197,39,107,100],[197,39,112,100],[213,39,94,110],[212,40,95,111],[211,41,97,111],[209,42,99,112],[209,43,98,112],[213,43,87,112],[213,42,83,113],[211,40,86,109],[211,38,86,103],[211,37,88,112],[211,20,186,131],[213,27,167,122],[212,44,87,105],[210,44,88,98],[195,44,106,98],[189,44,110,98],[182,46,117,99],[173,44,118,96,1],[161,43,130,99],[154,42,137,97],[153,42,137,97],[153,42,137,97],[152,41,137,98],[151,41,137,97],[149,41,145,97],[148,25,144,114],[148,13,144,126],[141,12,153,127],[115,11,173,128],[108,7,180,133],[108,4,180,136],[108,3,176,137,1],[108,1,161,139],[105,1,235,138],[103,1,295,148],[103,0,277,149],[108,0,234,137],[101,0,232,137],[99,0,135,139],[95,0,244,139],[81,0,152,139],[69,0,164,139,1],[66,0,169,139],[65,0,170,139],[63,0,168,138],[61,0,159,138],[35,0,304,139],[19,0,189,140],[18,11,138,129],[18,11,137,129],[18,11,137,128],[18,6,135,133],[7,4,146,136],[6,4,147,136],[3,4,150,136,1],[3,5,150,135],[3,8,150,132],[4,6,394,145],[12,6,388,145],[11,8,389,144],[11,8,387,144],[11,8,387,143,1],[10,8,113,131],[11,8,111,131],[10,9,112,130],[12,9,116,130],[12,9,111,130],[12,9,111,130],[12,9,110,131],[12,34,113,106],[13,35,110,104]],   //this function created all the DIV tags and adds them to the page every 83ms. (approximaterly 12 FPS) l=function(){ var a=d[f], c=document.getElementById("hplogo");  if(c&&a[0]){  var b=document.createElement("div");"hplogo"+f;[0]+"px";[1]+"px";[2]+"px";[3]+"px";"url(logos/2011/graham11-hp-sprite.png) no-repeat "+-g+"px "+-h+"px"; b.onmousedown=k; a[3]>i&&(i=a[3]); a[4]?(g=0,h+=i,i=0):g+=a[2]; c.appendChild(b); ++f; f< e && ( j=window.setTimeout(l,83) ) } }
I hope I have succeded to shine some light on how this Google doodle was made.


Popular posts from this blog

Basic cell counting and segmentation in Matlab

Counting cells manually is a tedious error prone process for humans. Given a large data set of microscopy images this task can be achieved much faster by means of basic computer vision techniques. In this tutorial we will segment cells from an image following a method similar to the one presented by Yongming Chen in 1999. The method uses basic morphological operations and the watershed algorithm to segment the cells. Nowadays better methods for cell segmentation exist. This method was chosen for its simplicity and ease of implementation.

We start with an image of cell-like structures by Anna-Katerina Hadjantonakis and Virginia E Papaioannou.

A = imread('cells.jpg');
We convert the image to grayscale:
I = rgb2gray(A);

To be able to extract the dimmer cells, it is necessary to perform some local contrast adjustments
I = adapthisteq(I);

Objects on the borders can be caused by noise and other artifacts. We can eliminate objects on the borders of the image like this:
I = imclearborder(…

Project planning in a text file

Whenever you work on a project it is important to be able to plan it ahead of time. This holds true for small and big project, from planning a trip to the spa to building a spaceship. The small project plans can be maintained in you thoughts while bigger ones require tools to help you see the big-picture of the project and manage task at a lower level. There are projects which start with a fully prepared plan and projects which pivot overnight, thus invalidating any original plan. For the latter flexibility is very important, and tools like Trello offer a great solution because they can be adjusted to fit your project.

However, it may happen sometimes that the project starts adjusting to the tool or that you still want to maintain a bigger picture of the main points of the project. You may also need to produce a rough development schedule to serve as a long term road-map.

I have prototyped a tool (and defined a workflow) which allows you to plan such projects.

To better understand how…

Selenium testing in Jenkins with an in-memory X server

This articles explains how to setup integration testing of web applications (that require a browser instance) on a Jenkins server. The article assumes you are able to understand the title. If you need a reminder follow the links below.

What is Selenium? Read here.
What is Jenkins? Read here.
What is an X server? Read here.

If you are still puzzled, this article is not for you.

Integration testing of web application is more complicated than unit testing, because they require a browser instance to be running. In order to run a browser (such as Firefox) you need and X server. The testing server usually doesn't have one.

One solution to this is to run the browser instance on another machine through VNC. This has the advantage of not using the same resources as the testing server. However it requires setting up a new machine, and configuring the integration tests to run a browser in it.

Another solution is to use Xvfb, a display server that implements the X11 protocol, but does not re…