Comments on: SSAO in stage 3D (source added)

By: Jason Huang

Jason Huang — Tue, 15 Jan 2013 18:29:38 +0000

Hi, I think the position and normal buffer should be in the world view space.
Before the flash 11.7 comes out, there is no such a thing called “floating texture”.
So the only thing you can do for the buffers is reconstructing them in the SSAO shader.
which are useing the quad uv.xy and the depth buffer and an interted projection matrix for
the position buffer, Normal buffer is much simple though, simply * 2-1 which means make it
back to range -1 ~ 1;
I will be finishing the SSAO using Flare3d framework today, adding a blur pass and removing
the artifacts as much as possible:)

By: _kihu

_kihu — Mon, 14 Jan 2013 11:43:46 +0000

Thank you for the answer! I’m still struggling to get this to work but your code is very helpful, and it’s also – I think – the only example of SSAO written in AS3. If I ever finish this, I’ll share the results and insights:)

By: bwhiting

bwhiting — Thu, 10 Jan 2013 09:02:02 +0000

At the moment you have to render the scene to a texture and output the depth to that, for any sort of usable accuracy you will need to encode it across more than one channel.

So in a nutshell, set render to a texture. and render your scene with a shader that outputs the depth. You then use this texture to extract the depth from in any subsequent passes.

By: _kihu

_kihu — Thu, 10 Jan 2013 08:37:02 +0000

Hey, very nice research!

How do you acquire the depth map for your calculations? As in tex_depth in lines below:

//sample deopth at current fragment, and decode
AGAL.tex(colour, uv_in, tex_depth, “2d”, “clamp”, flags);//ex ft0, v0, fs0 \n”+
AGAL.decodeFloatFromRGBA(depth, colour, depth_decoder);

I can’t find a way to get to Stage3D’s depth buffer.

By: GoroMatsumoto

GoroMatsumoto — Sat, 27 Oct 2012 19:19:04 +0000

Thanks a lot!

Your post is always impressive for me.
I will try again with ur helper class.

By: bwhiting

bwhiting — Wed, 24 Oct 2012 12:41:18 +0000

So far I have always been using my original AGAL helper, just haven’t updated the google code version for a while.

You only need to decode if you have encoded the depth across all the channels (rgba), if you have depth stored in just one channel (or the same in all channels) then you can get the value right back out easily (but the precision will not be as good).

the code for the decode is very simple:
AGAL.dp4(dest, rgba, constants); //where constants is [1/(256*256*256), 1/(256*256), 1/256, 1] – as long as this works with your chosen encode method

hope that helps, will update the AGAL helper again soon.

By: GoroMatsumoto

GoroMatsumoto — Wed, 24 Oct 2012 12:29:27 +0000

Hi. Thank you for sharing nice article : )

Which class do you use as AGAL helper?
asgl? or your original helper class?

I can’t find “decode” method in each of them.

Thank you in advance!

By: bwhiting

bwhiting — Fri, 22 Jun 2012 08:23:18 +0000

Yes I will do! It is by no means complete but I would be happy to release it in its current state. Someone else might be able to tweak it and get it working even better!

It’s all simple AGAL code at the moment and not that much of it either so should be easy to plug into something else.

If I get some time today I’ll put up a quick interactive demo and attach the shader source.

By: @Seraf_NSS

@Seraf_NSS — Fri, 22 Jun 2012 08:00:43 +0000

Great !
do you think about release the code source ?

By: bwhiting

bwhiting — Mon, 18 Jun 2012 08:16:01 +0000

Most of my models I get from:

http://www.turbosquid.com/
http://artist-3d.com/
http://archive3d.net/
and lots of cg forums etc…

Am not entirely sure about the batman one although I think I found a link to him on youtube!