I'm using an image in a canvas element as a texture in Three.js, performing image manipulations on the canvas using JavaScript, and then calling needsUpdate() on the texture. This works, but it's quite slow.
I'd like to perform the image calculations in a fragment shader instead. I've found many examples which almost do this:
Shader materials: http://mrdoob.github.io/three.js/examples/webgl_shader2.html This example shows image manipulations performed in a fragment shader, but that shader is functioning as the fragment shader of an entire material. I only want to use the shader on a texture, and then use the texture as a component of a second material.
Render to texture: https://threejsdoc.appspot.com/doc/three.js/examples/webgl_rtt.html This shows rendering the entire scene to a WebGLRenderTarget and using that as the texture in a material. I only want to pre-process an image, not render an entire scene.
Effects composer: http://www.airtightinteractive.com/demos/js/shaders/preview/ This shows applying shaders as a post-process to the entire scene.
Edit: Here's another one:
Render to another scene: http://relicweb.com/webgl/rt.html This example, referenced in Three.js Retrieve data from WebGLRenderTarget (water sim), uses a second scene with its own orthographic camera to render a dynamic texture to a WebGLRenderTarget, which is then used as a texture in the primary scene. I guess this is a special case of the first "render to texture" example listed above, and would probably work for me, but seems over-complicated.
As I understand it, ideally I'd be able to make a new framebuffer object with its own fragment shader, render it on its own, and use its output as a texture uniform for another material's fragment shader. Is this possible?
Edit 2: It looks like I might be asking something similar to this: Shader Materials and GL Framebuffers in THREE.js ...though the question doesn't appear to have been resolved.
Render to texture and Render to another scene as listed above are the same thing, and are the technique you want. To explain:
In vanilla WebGL the way you do this kind of thing is by creating a framebuffer object (FBO) from scratch, binding a texture to it, and rendering it with the shader of your choice. Concepts like "scene" and "camera" aren't involved, and it's kind of a complicated process. Here's an example:
http://learningwebgl.com/blog/?p=1786
But this also happens to be essentially what Three.js does when you use it to render a scene with a camera: the renderer outputs to a framebuffer, which in its basic usage goes straight to the screen. So if you instruct it to render to a new WebGLRenderTarget instead, you can use whatever the camera sees as the input texture of a second material. All the complicated stuff is still happening, but behind the scenes, which is the beauty of Three.js. :)
So: To replicate a WebGL setup of an FBO containing a single rendered texture, as mentioned in the comments, just make a new scene containing an orthographic camera and a single plane with a material using the desired texture, then render to a new WebGLRenderTarget using your custom shader:
// new render-to-texture scene
myScene = new THREE.Scene();
// you may need to modify these parameters
var renderTargetParams = {
minFilter:THREE.LinearFilter,
stencilBuffer:false,
depthBuffer:false
};
myImage = THREE.ImageUtils.loadTexture( 'path/to/texture.png',
new THREE.UVMapping(), function() { myCallbackFunction(); } );
imageWidth = myImage.image.width;
imageHeight = myImage.image.height;
// create buffer
myTexture = new THREE.WebGLRenderTarget( width, height, renderTargetParams );
// custom RTT materials
myUniforms = {
colorMap: { type: "t", value: myImage },
};
myTextureMat = new THREE.ShaderMaterial({
uniforms: myUniforms,
vertexShader: document.getElementById( 'my_custom_vs' ).textContent,
fragmentShader: document.getElementById( 'my_custom_fs' ).textContent
});
// Setup render-to-texture scene
myCamera = new THREE.OrthographicCamera( imageWidth / - 2,
imageWidth / 2,
imageHeight / 2,
imageHeight / - 2, -10000, 10000 );
var myTextureGeo = new THREE.PlaneGeometry( imageWidth, imageHeight );
myTextureMesh = new THREE.Mesh( myTextureGeo, myTextureMat );
myTextureMesh.position.z = -100;
myScene.add( myTextureMesh );
renderer.render( myScene, myCamera, myTexture, true );
Once you've rendered the new scene, myTexture will be available for use as a texture in another material in your main scene. Note that you may want to trigger the first render with the callback function in the loadTexture() call, so that it won't try to render until the source image has loaded.
Related
I am trying to apply an image dynamically on a gltf loaded mesh.
The code to load the model looks like:
const gltfLoader = new GLTFLoader();
const url = 'resources/models/mesh.gltf';
gltfLoader.load(url, (gltf) => {
const root = gltf.scene;
scene.add(root);
})
When looking from top the element looks like a rounded rect:
When inspecting the imported mesh I can see that the BufferGeometry has a count of 18.000 points:
Everything works fine however if I apply the texture like this:
const texture = new THREE.TextureLoader().load( 'textures/land_ocean_ice_cloud_2048.jpg' );
const material = new THREE.MeshBasicMaterial( { map: texture } );
root.children[0].material = material;
The image is not visible but the mesh is now colored in 1 color.
Is it possible to apply the image just on the top face of the rect?
Hard to tell what the problem is without seeing the resulting image. However, I would just assign a new texture like this: root.children[0].material.map = texture instead of creating a whole new material, since you don't want to lose all the material attributes that came in the GLTF.
Additionally, MeshBasicMaterial always looks flat because it is not affected by lights.
I have a mesh that I load as a .dae (collada). However, my texture file is separate (as a PNG)
Currently, my loading code is
loader.load("./assets/map_tutorial.dae", function(collada) {
var terrain = collada.scene.children[0];
var texture = new THREE.ImageUtils.loadTexture("./assets/map_tutorial_tex.png");
terrain.rotation.x = Math.PI / 2;
terrain.rotation.y = Math.PI;
scene.add(terrain);
console.log("There are " + collada.scene.children.length + " meshes!");
});
However, I'm uncertain as to how to apply the texture to my mesh (terrain)
Thanks in advance!
You will definitely have to create new material object from your texture. Assuming you want lambert, it would be like this:
var material = new THREE.MeshLambertMaterial( { map:texture } );
I am not familiar with collada loader, but it seems that it already created a mesh for you. In this case I am not sure whether you can change material of this mesh. What should definitely work is to create new mesh from geometry that you loaded from .dae file and material that you created from png image.
var mesh = new THREE.Mesh( terrain.geometry, material );
scene.add( mesh );
I hope this helps, normally I use three.js native JSON model files where this approach is working. Difference is that THREE.JSONLoader gives you two objects for your usage: geometry and materials. Collada loader seems to provide you with already created mesh so try to change it's geometry if possible or "steal" the geometry from this mesh and create a new one like I did in the example.
This is my code:
var sprite = new THREE.Sprite(material);
sprite.renderDepth = 10;
The above renderDepth setting is invalid, it does not work for sprites.
How to solve this problem?
You want one sprite to always be on top.
Since SpriteMaterial does not support a user-specified renderDepth, you have to implement a work-around.
Sprites are rendered last when using WebGLRenderer.
The easiest way to do what you want is to have two scenes and two render passes, with one sprite in the second scene like so:
renderer.autoClear = false;
scene2.add( sprite2 );
then in the render loop
renderer.render( scene, camera );
renderer.clearDepth();
renderer.render( scene2, camera );
three.js r.64
I create a Mesh having a PlaneGeometry and the material defined by a texture loaded from a JPEG image. Everything is fine, excepting that there is a small amount of time before the texture image is loaded when the plane is displayed using a dark color. Is there a way to change this color to something else?
I tried the color option for material, but it is not applied.
var texture = new THREE.ImageUtils.loadTexture('/path/to/image');
texture.minFilter = THREE.LinearMipMapLinearFilter;
texture.magFilter = THREE.NearestFilter;
var material = new THREE.MeshBasicMaterial({
side : THREE.DoubleSide,
map : texture,
color : 0xf0f0f0
// this doesn't seem to work
});
var geometry = new THREE.PlaneGeometry(Math.abs(line.x1 - line.x0), depth);
var mesh = new THREE.Mesh(geometry, material);
That black color is the texture rendering without any texture data. The easiest fix is to load the texture and the mesh, but do not render the mesh until both have fully loaded.
Another option is to create a very small 1x1 texture that is the color you want, use that as your texture initially, and then change the mesh material to your final texture once the desired texture has fully loaded.
have a 3d maze with walls and floor.
have an image with a key ( or other object its not important, but all of em are images and not 3d models ).
I want to display it on the floor and if the camera moves around the object needs to look the same without rotating the object. How can i achieve this?
Update1:
I created a plane geometry added the image ( its a transparent png ) and rotating at render. Its working good, but if i turn the camera sometimes the plane lose transparency for about a few milisec and the get a solid black background ( blinking ).
Any idea why?
here is the code:
var texture = new THREE.ImageUtils.loadTexture('assets/images/sign.png');
var material = new THREE.MeshBasicMaterial( {map: texture, transparent: true} );
plane = new THREE.Mesh(new THREE.PlaneGeometry(115, 115,1,1), material );
plane.position.set(500, 0, 1500);
scene.add(plane);
// at render:
plane.rotation.copy( camera.rotation );
This will be achieved by using:
function animate() {
not3dObject.rotation.z = camera.rotation.z;
not3dObject.rotation.x = camera.rotation.x;
not3dObject.rotation.y = camera.rotation.y;
...
render();
}