Surfin' Safari

CSS Animation

Posted by Dave Hyatt on Wednesday, October 31st, 2007 at 1:06 am

We have another cool new CSS feature to talk about: animation specified in CSS. There is a lot of ground to cover here, so we’ll start with the basics first.

The simplest kind of animation that we’ve added support for is called a transition. Normally when the value of a CSS property changes, the rendered result is instantly updated, with the element in question changing immediately from the old property value to the new property value. Transitions describe how to instead execute an animation from the old state to the new state over time.

Transitions are specified using the following properties:

transition-property – What property should animate, e.g., opacity.
transition-duration – How long the transition should last.
transition-timing-function – The timing function for the transition (e.g., linear vs. ease-in vs. a custom cubic bezier function).
transition – A shorthand for all three properties.

Here is a simple example:

div {
  opacity: 1;
  -webkit-transition: opacity 1s linear;

div:hover {
  opacity: 0;
This div will fade out when hovered over. In browsers that do not support this animation, there will be graceful degradation, as the div will simply fade out immediately.

Each of these properties supports a comma separated list of values, like CSS3 multiple backgrounds, which allows different transitions for individual properties to be described in a single style rule. Each value in the list corresponds to the value at that same position in the other properties.

For example:

div {
  -webkit-transition-property: opacity, left;
  -webkit-transition-duration: 2s, 4s;

In the above style rule, opacity will animate over 2 seconds, but left will animate over 4 seconds.

Some very complex properties can actually be animated. Take, for example, the new -webkit-transform property. Here’s an example of a simple spin effect using -webkit-transform and -webkit-transition.

<div style="-webkit-transition: -webkit-transform 3s ease-in" 
This div will do a spin when clicked!
This div will do a spin the first time it is clicked!

Borders can also be animated. The following box has a simple border animation where the border will both grow in thickness and change color when the box is hovered.

This div will acquire a slightly thicker blue border when hovered.

Note with the hovering examples that the animation will reverse itself when the mouse moves out of the div. Any time the property changes value, the animation will simply start again with the current position as the from value and the new value as the destination. The transition properties of the source state are used to figure out how to animate to the new target state.

The key points to understand about transitions are:
(1) They are implicit animations. Scripts and stylesheets can be written as normal, and the animations will simply happen implicitly as the properties change values.
(2) They degrade gracefully. Browsers that do not support transitions simply won’t animate, but otherwise all code and style rules can remain the same.

Here are more detailed descriptions of the properties. All of these properties can take multiple comma-separated values.

Values: none | all | <property>
Initial Value: all
Description: Specifies what property triggers an animation. A value of none means there is no transition. A value of all means that every animatable property triggers an animation. Otherwise an animation will only trigger when the exact specified property changes value.

Values: <time>
Inital Value: 0
Description: Specifies how long the transition should take. The default is 0.

Values: ease | linear | ease-in | ease-out | ease-in-out | cubic-bezier(x1, y1, x2, y2)
Initial Value: ease
Description: The transition-timing-function property describes how the animation will proceed over time. Keywords can be used for common functions or the control points for a cubic bezier function can be given for complete control of the transition function. To specify a cubic bezier function, you give an X and Y values for 2 of the control points of the curve. Point P0 is implicitly set to (0,0) and P3 is implicitly set to (1,1). These 4 points are used to compute a cubic bezier curve.

The timing function keywords have the following definitions:
linear – The linear function just returns as its output the input that it received.
defaultease – The default function, ease, is equivalent to cubic-bezier(0.25, 0.1, 0.25, 1.0).
ease-in – The ease-in function is equivalent to cubic-bezier(0.42, 0, 1.0, 1.0).
ease-out – The ease-out function is equivalent to cubic-bezier(0, 0, 0.58, 1.0).
ease-in-out – The ease-in-out function is equivalent to cubic-bezier(0.42, 0, 0.58, 1.0)
cubic-bezier – Specifies a cubic-bezier curve whose P0 and P3 points are (0,0) and (1,1) respectively. The four values specify points P1 and P2 of the curve as (x1, y1, x2, y2).

In future posts I’ll go into transitions in more detail and also talk about another feature we’re adding: explicit animations. We are also preparing a more detailed proposal (full of intimidating spec language) that covers our thoughts on transforms, animation and other advanced visual effects.

[1 Dec 2008 – updated timing function values to reflect new implementation. ‘default’ is now called ‘ease’. Refer to CSS Transitions specification for details.]

37 Responses to “CSS Animation”

  1. ChrisJay Says:

    Truly, truly awesome. Together with client-side storage and transformations, you are advancing the web in ways not seen for years – well done.
    Does this constitute a “native SDK” for the iPhone? :)
    One more point – here’s hoping that your explicit animations will be in the functional programming style, so that e.g. div1 width can depend in real-time on div2 height.

  2. theallan Says:

    Very interesting – and excellent work guys! The animations are very smooth indeed, unlike a lot of javascript animations. Declarative animation in CSS is quite a unique idea I believe, and I’m wondering where you see this in relationship to declarative animation in SVG (which I don’t believe Webkit supports yet)? Will progress in this area of SVG be halted now that you have these CSS attributes, or do you see them as completely separate threads?

  3. Antoine Quint Says:

    I have the same question as Theallan regarding future work on declarative SVG animation. It would be a real shame in my opinion to see this in lieu of the generic, multi-purpose declarative animation support in SVG, which is already well supported in Opera for a few versions now. I understand it must be much easier in terms of development cost to extend CSS to support these features, but I’m still hopeful for more SVG animation goodies in WebKit.

  4. Pingback from Webkit introduce more new features - CSS3 . Info:

    […] Animation, which uses the -webkit-transition family of properties, allows you to set timings for fades, rotation, expansion, collapses, and more. They work in the same way as many current JavaScript libraries. […]

  5. maciej Says:

    Antoine, we’re not against SVG animations, and that feature will likely come eventually. For HTML, however, it makes more sense for animations to be in the presentational layer, CSS. On another note, there’s also no reason you can’t apply CSS animations to SVG.

  6. cheeaun Says:

    Wow, this is cool. Few questions though:
    1. ‘transition-property’ only accepts one property? So, to apply animation specifically to two properties, two ‘transition-property’s are needed? The ‘all’ value seems a little too much 😀
    2. No time intervals? I mean, periodic animations?

  7. braun Says:

    Great, it looks cool to me as well :).
    My question is if it is possible to specify different transitions for different pseudoclasses (“events”)? Eg. is possible to specify one transition for the hover and the other for the active state? Currently it’s hard to spot from the CSS rules when the transition is going to fire.

  8. Pingback from CSS Animation « outaTiME:

    […] Surfin’ Safari.) Posted by outaTiME Filed in CSS, Development, […]

  9. asdasdwr Says:

    Tricky question: how do you animate change of display property? or creation/removal of generated content?
    In my stylesheets these are two most common declarative “effects”.

  10. xenon Says:

    @cheeaun, all of the transition CSS “properties supports a comma separated list of values, like CSS3 multiple backgrounds, which allows different transitions for individual properties to be described in a single style rule.”

  11. hyatt Says:

    This is independent of SVG animation, which is still being worked on. Basically we feel that in order to add simple animations to a Web site, one shouldn’t have to convert the entire Web site to a compound document with SVG+XHTML just to animate something. SVG+XHTML also doesn’t degrade in older browsers (or in Internet Explorer).

  12. hyatt Says:

    braun, the transition properties set on the old state determines what happens when you change to a new state. This means that you can change the transition properties if you want a new animation. So for example you could have

    div {
    transition: all 5s ease-in;


    div:hover {
    transition: all 10s ease-out;
    opacity 0

    In the above example when you hovered over the div, you’d get a 5 second fade-out. If you move the mouse off the object however you get a 10 second fade-in.

  13. hyatt Says:

    cheeaun, transitions are for basic implicit animations. We will also be introducing explicit animations (using actual “animation” CSS properties). I’m going to blog about that more once there’s an actual prototype to play with checked in to the tree.

  14. Pingback from » Blog Archive » Declarative CSS, the death of Scriptaculous?:

    […] 31st, 2007 in Links Examples of declarative CSS transitions and animations, which are available in recent versions of Safari. Many of the Scriptaculous features are more […]

  15. caesarsgrunt Says:

    But… are these properties a part of any (present or future) spec? Otherwise, whilst I certainly like them, it seems a bit sort of browser-wars-ish implementing proprietary features like this…
    But good work anyway!

  16. hyatt Says:

    Yes, we plan to propose these for inclusion in the CSS spec.

  17. maciej Says:

    Keep in mind, WebKit nightlies are not a product release, they are an early access prototype. We do plan to take this stuff to standards.

  18. Grauw Says:

    I proposed transition effects on the CSS WG list in May last year:

    Although the responses were a bit tame, I thought it was a really nice idea, and I’m glad to see this taken up by a browser vendor (be it because they read my message to the list or came up with it independently ;)). Hope that now this is backed by Safari, you will be able to get this standardised!


  19. Grauw Says:

    (it’s suspiciously similar to my proposed syntax, hehe ;p)

  20. Grauw Says:

    By the way, I’d say for implicit animations use CSS (because it’s so easy and fits so nicely with the rest of the stylesheet), and for explicit ones use SMIL.

  21. hyatt Says:

    Grauw, yeah that is really similar. I’ve never read your message either! That validates the approach in my opinion. :)

  22. pixelbath Says:

    So, what happens if these are not included in the CSS specification? Will you cease working on proprietary CSS extensions, or go the way of Microsoft with DirectShow filters and page transitions?

    I see some cool things, and hope you might lead the charge on making CSS3 usable, but I just hope we don’t end up with more proprietary baggage on the web.

  23. hyatt Says:

    Like Microsoft our engine is also used inside many applications on our home operating system. We serve two masters: the Web and OS X (in all its forms). If people need these capabilities on OS X, we are going to provide them, regardless of whether or not they end up in any specification. That said, we are very interested in standardizing this stuff, and are preparing a proposal for the folks at the CSS WG.

  24. hyatt Says:

    Also, CSS has a well-documented means of providing safe non-conflicting extensions (using vendor-specific prefixes), so there is absolutely nothing wrong with CSS extensions. This isn’t like HTML, where you end up polluting the namespace. CSS can actually be extended safely in a way that doesn’t conflict with the standards.

  25. bstewart Says:

    Is it appropriate that Safari version 3.0.3 (522.15.5) for Windows doesn’t show the animation effects? I was hoping to get to see them and the examples are behaving as stated for unsupported browsers. I’m assuming that Safari for Mac work properly. Is that correct?

  26. Mark Rowe Says:

    These features are only available in nightly builds of WebKit at this point. Visit and grab the latest Windows build to test it out.

  27. bstewart Says:

    @Mark Rowe
    Thanks for the speedy reply and point in the right direction.

  28. bstewart Says:

    FYI, it doesn’t appear to be functioning as described in Safari for Windows using Vista Home Basic. :)

  29. Mark Rowe Says:

    If something is not working correctly you should file a bug report. Be sure to double-check that you are in fact testing with a nightly build.

  30. Jakob Peterhänsel Says:

    Wow, way cool.

    One thing that pops up in my mind, without having made a test other that with this page:

    Does the element in question (like, in a fade-out) still receive events? It seems so on example 3, and that would be cool when used on sub-menues, so they can start a fadeout when the cursor leaves the ‘safe-area’ and give visual feedback, instead of just popping away. Would make navigation more safe, I guess… :-)

  31. Brad Says:

    In the first example, it seems to fade in much quicker than it fades out. Is that just an illusion?

    BTW, I think this is fantastic work!

  32. hyatt Says:

    Jakob, yes the element will still receive events.

  33. caesarsgrunt Says:

    I understand what you’re saying about providing these features for OS X’s sake even if they don’t get standardised, and in principle I think that’s great (though of course, I really hope they do get standardised). However, I’ve got a sort of endless loop-style argument going round inside my head about whether it’s ‘safe’ or not…
    The main problem that I see is the -webkit-. I fully understand what this is for (although I can’t remember what it’s called) and to be honest, I can’t think of a better alternative – but I still don’t like it! The same applies to -moz- and -o-, of course. I end up having to specify things like opacity four times (including Microsoft’s stupid filter) in order to make it work cross-(modern)-browser!

    Can anyone tell me the difference between -khtml- and -webkit-? I understand that their roots are obviously different, but in my limited experiments both seem to function identically. I tend to use -khtml-, mainly because it sounds like it’s probably more likely to be compatible with KHTML browsers – are the two functionally identical in WebKit? Or perhaps in both WebKit and KHTML…? Which should I use? (I’m not sure if there’s even a difference between WebKit and KHTML anymore… Is there?)

    On a subject slightly more relevant to the original post – is there any way to use these properties to achieve transitions between pages, like with Microsoft’s horrible proprietary code?

    BTW – I like this (experimental, I know) feature so much that I’m already using it on all my sites. :) (To smooth off the sudden switches in link style when hovered, for example.)

    Keep up the good work!

  34. caesarsgrunt Says:

    Another question – sorry!
    Is there a reference of ‘experimental’ -webkit- css anywhere, other than scattered through this blog? It would be great if there was – sort of like Mozilla have, but hopefully more detailed :)

  35. Pingback from Surfin’ Safari - Blog Archive » CSS Animation at Nuclear Blog:

    […] Surfin’ Safari – Blog Archive » CSS Animation We have another cool new CSS feature to talk about: animation specified in CSS. […]

  36. pohl Says:

    What do I need to view this page in in order to see these effects for myself? It appears that
    I’m only seeing the degraded behaviour.

    I have tried Safari 3.0.3 (522.12.1) on Tiger, and
    Safari 3.0.4 (5523.10) on Leopard.

  37. Mark Rowe Says:

    As previously mentioned:

    These features are only available in nightly builds of WebKit at this point. Visit and grab the latest build to try them out.