Im building an iOS AIR app using AS3/Flash builder.
The app grabs the camera stream into a video object and then draws that to a BitmapData. The camera stream is 1280x720px and the container is 2208x1242px (iPhone 7+) so I need to scale the footage. I also need to rotate it around the center point:
mat.identity();
mat.translate(-video.width/2,-video.height/2);
mat.rotate(Math.PI/2);
mat.translate(video.height/2,video.width/2);
mat.scale(deviceSize.height/cam.width,deviceSize.width/cam.height);
I have timed the drawing operation with matrix:
videoBitmapData.draw(video,mat); //9-11ms
And without matrix:
videoBitmapData.draw(video); //2-3ms
Clearly the transformation is slowing me down.
Is there a faster way to do this? Maybe drawing first, then applying the matrix somehow?
Pre-rotating something?
Leveraging native scaling/rotation?
Can I skip using the video object somehow? Accessing the camera data in a more raw form?
I see no difference using render mode GPU vs CPU.
Thanks!
Edit:
I managed to get down to about half by doing this:
1. Put the video object inside a Sprite
2. Transform the Sprite.
3. Draw the Sprite to the BitmapData.
(5-6ms)
Suprisingly, this was more consistent than this:
1. Put the video object inside a Sprite
2. Transform the video object inside the Sprite
3. Draw the sprite to the BitmapData
(5-12ms)
Bonus:
I only needed part of the pixels from the camera (full width x 100px of the height). When realizing I could use clipRect with the .draw command I managed to get down to 0-1ms.
Thanks to Organis
Related
I am making a game in AS3 for a school project (using the AIR api). I have about a year's worth of experience in AS3, so i would say i'm proficient but not an expert. Anyway, i have never attempted to do sprite animations in AS3 and i'm not quite sure how to approach it.
If i create a list of bitmaps and call addChild() and removeChild() to display each frame of the animation, it would affect the framerate as these functions are not very efficient (i have tried this before and it tanked my framerate if i had too many animations going at once). I have also tried creating a list of BitmapData objects, adding a bitmap to the display list, and then pointing it to a different BitmapData each frame, but this does not seem to work at all.
So what is the best way to do this? In XNA, for example, i would create a sprite class that draws to the screen using a sprite batch, then i would create a list of sprite objects and cycle through them to create the animation. Is there any way to achieve a similar result in actionscript?
First (simple) method: you can just use multi-frame MovieClip for animation. Put an image in each frame, put a MovieClip on the stage and that's all. You can control this animation using play(), stop(), gotoAndPlay(), gotoAndStop(). It will work without much problems for a generic platform game (I did that myself long ago).
Second (advanced) method: use bitmap blitting. For each animation, create a bitmap image that holds each frame of the animation. Use this image as a source for copying pixels into your current animated object. You just need to copy a particular rectangle area inside a source bitmap that corresponds to the current frame.
The actual blitting happens here
destinationBitmapData.copyPixels(sourceBitmapData, areaRectangle, destinationPoint);
Where destinationBitmapData is your "canvas" that you're blitting to; sourceBitmapData is the source image holding all animation frames; areaRectangle is the Rectangle inside the source image defining which area to copy; destinationPoint is left-top coordinate of the copy area in your canvas.
The destination canvas can be just one small object (like your game character that is moving around) or the entire game screen with all objects. I.e. instead of blitting and adding each object separately, you can just have one big canvas and blit any necessary parts directly to it.
That said, there is already a number of various ready-made engines that use blitting and even advanced techniques like 3D acceleration for 2D sprites.
One of them is Starling.
In my flash project I have 3 layers.
back layer contain Flvplayback video
middle layer contain webcam stream
front layer contain flvplayback video(with Transparent background).
webcam stream video is also transparent.
I add the stream from following code.
webCam=Camera.getCamera();
webCam.setMode(appWidth/scale,appHeight/scale,60);
video=new Video(appWidth/scale,appHeight/scale);
video.smoothing=true;
video.attachCamera(webCam);
bmpdata=new BitmapData(appWidth/scale,appHeight/scale);
bmpdata_bg=new BitmapData(appWidth/scale,appHeight/scale);
bmp=new Bitmap(bmpdata);
bmp.smoothing=true;
bmp.x = 256;
bmp.y = 374;
addChild(bmp);
and I update that webcam stream using following function
function onFrameEnter(evt:Event):void {
bmpdata.draw(video);
}
I want to know how to add this webcam stream to a middle layer and play it.
Is there any way to add this webcam stream to flvplayback in the middle layer?
As far as I know it's impossible to refer to Flash IDE layers from Actionscript. There are a couple of workarounds you can do, though.
One is to have an empty instance movieclip in the desired layer with an instance name mcStream. Then, you can use mcStream.addChild(bmp);.
Another way is to refer to the item you want it to be under (we'll call it mcOverlay) and add it at its index. addChildAt(bmp, getChildIndex(mcOverlay)) This will push add it where the overlay is and push the overlay on top of it.
I definitely recommend having a look at http://www.adobe.com/devnet/flash/quickstart/display_list_programming_as3.html
this is a newer version to a question i have asked before but have not received an answer to.
I am developing a mobile AIR app with many animations and tests have shown that using bitmaps and GPU rendering give the best performance across all mobile models. I receive normal vector animations from the app's artists, and have built a system which loops through these animations at runtime and draws the content into bitmapdatas.
my concern is looping through the movieclip's frames. If I have these lines of code:
for (var i:uint=1; i<mc.totalFrames+1; i++) {
mc.gotoAndStop(i);
bitmapData.draw(mc);
}
I can't be sure the frame got "constructed" before being drawn, and my tests with an Android tablet prove this right - sometimes frames aren't drawn.
This mc is off the display list obviously (we dont need to render to the screen). So is there a way to make sure the frame has been built before drawing it to a bitmapdata? (and waiting for FRAME_CONSTRUCTED, EXIT_FRAME, etc.. is obviously slow and unneeded)
any ideas?
thanx
Saar
more info to clarify:
this is a children's book. each page has animations in it. all these animations are vector animations in timelines in FLAs I receive from the devloping artists (not the users).
Every page has an FLA which i publish to a swf.
what i actually do is replace vector animations in loaded SWFs with bitmap version of them.
At app runtime, upon each new page, i load the swf, go though all the animations in it (looping through the content's children) and each movieclip i rasterize into array of bitmapdatas.
My custom "Bitmap Movieclip" places on the displaylist a replcaement bitmap for each movieclip, and on ENTER_FRAME i switch the bitmaps' bitmapdatas from the array.
this gives very hight performance
I'd suggest use a splash screen, and place your MC to be converted to the bitmap to the stage first, then draw to your bitmapData. Several issues had me force to put a DisplayObject being drawn to the stage before actual drawing once already, so this might be a quick fix for you too.
var splashBD:BitmapData=new BitmapData(stage.stageWidth,stage.stageHeight,false,0xffffff);
var splashBM:Bitmap=new Bitmap(splashBD); // draw a thing as needed here
addChild(splashBM);
addChildAt(MC_to_convert,getChildIndex(splashBM));
// then advance it by frames and draw.
sorry MY BAD
waiting for FRAME_CONSTRUCTED is the answer
I'm making an animation of a 2D character that can walk, run, jump, bend,...
Would it be better to load one big 'spritesheet' with all the animations and just use a mask, or would loading separate files (walk, run,...) be better because you're not using a mask on such a big image every frame?
I'm not using the Stage3D features with a framework like Starling because I think the normal flash display API is fast enough and has much less bugs than the relatively new GPU frameworks.
Blitting just the character (using lock(),copyPixels(),unlock()) works pretty well.
private function updatePixels():void{
//update sprite sheet copy position based on the frame placements ons prite sheet
position.x = spriteSourceData[currentFrame].x + offset.x;
position.y = spriteSourceData[currentFrame].y + offset.y;
//draw into the bitmap displayed
displayData.lock();
displayData.fillRect(displayData.rect, 0x00FFFFFF);//clear
displayData.copyPixels(sourceData, spriteData[currentFrame], position);//copy new frame pixels
displayData.unlock();
}
//a bit about vars:
position:Point
spriteSourceData:Vector.<Rectangle> - from parsed Texture Packer data
offset:Point - front view and side view animations weren't always centred, so an offset was needed
displayData:BitmapData - pluging into a Bitmap object displayed
sourceData:BitmapData - the large sprite sheet
currentFrame:int - image index on the sprite sheet
I've done this on an older project writing a custom class loosely following what I've learned from Lee Brimelow's tutorial series Sprite Sheets and Blitting (Part 1,Part 2, Part 3)
In short, you'd use two BitmapData objects:
a large sprite sheet
a small image to display just the character (size of the largest character bounding box) to copy pixels into
In my project I had a character with front and side animations and for the sides I've used one set of animations and used the Matrix class to flip(scale and translate) the side animation accordingly. I've used TexturePacker to export the image sequence as a sprite sheet and the frame data as well as a JSON object. There is native JSON support now, so that's handy. Texture Packer isn't free but it's really worth the money (affordable, fast and does the job perfectly). I haven't used Flash CS6 yet but I imagine it's also possible to import your image sequence and export a spritesheet with the new feature.
In my experience the rule "the simpler the display list the better the performance" generally applies. Which means you should use the most specific display object that will do the job (don't use a Sprite when a Shape would be sufficient or favor Bitmaps over vectors where it makes sense).
The most extreme version of this is to only have one Bitmap display object on the stage and use copyPixels to draw all game objects into it every time you want to update the screen. It doesn't really matter what the source in the copyPixel call is, it could either be a large BitmapData acting as a sprite sheet or a small BitmapData objects representing a single frame in an animation. This method is really fast and you could easily have many hundreds of objects on screen at the same time. But using copyPixels means you can't scale or rotate a game object, so for those cases you would have to fall back to the much slower draw() method. Of course this single Bitmap method is not suitable for games where you need to attach mouse events to specfic objects in the game, but works well for shoot'em ups or platform games.
To answer your question, I think you will get better performance by using a single Bitmap display object to represent the player and a collection of BitmapData objects for all the animation frames. Then you can just change the Bitmap's bitmapData property to the frame you want to display. You can still load a large spritesheet png and then plit it up into a series of BitmapData objects during the initialization of the game.
I've built a photo booth app for an installation and I need to take a screen shot of the bitmap data with it's frame... I foolishly nested my objects all wrong way too early in the game so taking a picture of just the display object is moot at this juncture, and I dont have time to re-organize everything.
I know the encoder can output stills of the stage, but can that be a defined set of coords? I have an 800x600 region that needs to output, ignoring the rest of the stage.
I am playing with other options as well, but if there is anything that seems obvious, i would greatly appreaciate it!
You can get the whole stage as a bitmap data and then use the copypixels method to copy the region you need.
http://help.adobe.com/en_US/FlashPlatform/reference/actionscript/3/flash/display/BitmapData.html#copyPixels%28%29
Or you can use the draw method of BitmapData class to draw a display object into that bitmap data.
http://help.adobe.com/en_US/FlashPlatform/reference/actionscript/3/flash/display/BitmapData.html#draw%28%29