CSSVR is prototype implementation that uses the DOM and CSS to create VR experiences. It's intended to test the feasiblity of creating progresive VR experiences using media queries and CSS transforms, if browsers internally rendered DOM trees to VR hardware.
This is a simluation, not a full VR solution. It's only compatiable with VR headsets that hold a mobile device (e.g. Google Cardboard) and lacks some of the features expected from a full VR experience, such as lens distortion.
You can read more about the CSSVR project in the accompanying article on my website.
cssvr media type, allowing VR-specifc styles to be conditionally applied to a document using a media rule (i.e. @media cssvr {...}):hover pseudo-class when the user "looks" at an object.click event for an element when the user hovers over it for a predetermined time.Create the CSS for your VR demo and include them in your test page using either a <link> element with a cssvr media type:
<link rel="stylesheet" media="cssvr">or, inline the styles with:
<style>
@media cssvr {
/* CSS rules go here */
}
</style>This prototype doesn't contain a CSS parser so it won't fetch and process the content of remote CSS content. This means you can't use the cssvr media type in any linked stylesheets.
<body>
<div id="root">
<!-- page markup -->
</div>
<script src="https://github.com/keithclark/cssvr/raw/master/path/to/cssvr.min.js"></script>
<script>
var vrConfig = {
root: document.getElementById('root')
};
CSSVR.start(vrConfig);
</script>
</body>| Property | Type | Description |
|---|---|---|
root |
DOMElement | The scene root element |
fov |
Integer | Field-of-view value |
projection |
String | stereo or normal |
stereoPitch |
Number | Distance between the pupils of the users eyes (in px) |
cameraHeight |
Number | Distance between the ground and the users head (in px) |
An example:
CSSVR.start({
root: document.getElementById('root'),
fov: 70,
projection: 'stereo'
});During initialization a set of DOM nodes are created to facilitate styling and control of the viewport, camera and scene. A set of base rules are used to clip content and enable 3D rendering (setting perspective etc.). Finally, the page content is moved into the new structure and the browser renders our content in 3D. For example, assuming this was the original DOM structure:
<h1>My test content<h1>After the VR script has been enabled, the DOM will look like this:
<div class="vr">
<div class="vr__eye">
<div class="vr__viewport">
<div class="vr__camera">
<h1>My test content<h1>
</div>
</div>
<div class="vr__cursor"></div>
</div>
</div>If stereoscopic rendering is enabled the entire DOM is cloned for the right eye and inserted next to the original version. (It's ok if alarm bells are ringing — this is terribly inefficent! Remember, this is a proof of concept.)
<div class="vr vr--stereoscopic">
<!-- LEFT EYE -->
<div class="vr__eye vr__eye--left">
<div class="vr__viewport">
<div class="vr__camera">
<h1>My test content</h1>
</div>
</div>
<div class="vr__cursor"></div>
</div>
<!-- RIGHT EYE -->
<div class="vr__eye vr__eye--right" aria-hidden="true">
<div class="vr__viewport">
<div class="vr__camera">
<h1>My test content</h1>
</div>
</div>
<div class="vr__cursor"></div>
</div>
</div>Now that there are two copies of the content, the horizontal component of the perspective-origin is set for each viewport (x) to acheive the stereoscopic effect:
.vr__eye--left .viewport {
perspective-origin: calc(50% - x) 50%;
}
.vr__eye--right .viewport {
perspective-origin: calc(50% + x) 50%;
}To keep both viewports in sync, a MutataionObserver is used to apply any DOM changes made in the left viewport to the right clone.
Once CSSVR is running, the browser will ignore any styles declared in screen media blocks and only apply styles in the cssvr media block. When the simulator is stopped, the page will return to it's original state.
Head tracking is acheived by listening for the deviceorientation event. The Euler angles provided to the event handler are converted into heading, elevation and tilt and then applied to the vr__camera element using a transform, syncing the viewport(s) to orientation of the device.
:hoverThe :hover pseudo-class is simulated by casting a ray from the centre of the viewport into the scene. The vertex data for each element is computed and then checked to see if it intersects the ray. The closest intersecting element is then selected as the hover target.
During startup, the simulator (crudely) replaces all instances of the string ":hover" in stylesheets with ".vr__selected". When an element is deemed to have intersected the ray the class is set and the relevant styles are automatically applied.
:hover honours the pointer-events style rule, allowing selections to be ignored by setting the value to none.
Once a hover target has been determined, it's attribute list is checked for the existence of vr-action and, if it's found, a timer is started. If the user remains focused on the element until the timer finishes, a click event is fired. If the user moves away from the element, the timer is cancelled. In an ideal world the vr-action attribute wouldn't be needed, but there's no way to check if an element has event listeners bound to it, so we need to give the simulator a hint.
1) Clone this repo.
2) Install dependencies: npm install
3) Build the project with the watch task: gulp dev
4) Start editing...
gulp dev - builds the unminified JS libray to the /build/ folder and watches the filesystem for changes.gulp dist - builds the both the unminified and minified distribution files to the /dist/ folder.gulp publish - builds a standalone website form the /examples directory.If you wish to build with your own config, you can pass in a CONFIG env. var when you start gulp:
$ CONFIG=myconfig gulp publishThis will cause the build script to load config-myconfig.json, rather than config-dev.json.