[SItoA] Adjusting camera exposure


Arnold added support for per-camera exposure awhile ago:

Camera exposure control: All cameras, even custom cameras, have an exposure parameter that scales the pixel samples by 2 ** exposure . Increasing the exposure by 1.0, or one stop, produces an image twice as bright, while reducing the exposure by one stop produces an image half as bright. The default value is 0 for backwards compatibility. You can also control this in in kick with the new -e option.

In Softimage, you can access the exposure parameter though the User Options.

user-options-camera-exposure

Changing ASS files with the Arnold Python API


If you want to change something in existing ASS files, don’t write an ad-hoc script or your own parser for the ASS file. Use the Arnold API. The Arnold API includes a set of Python bindings, so you can fairly quickly whip up a script to do whatever it is you need to do 🙂

For example, we recently discovered (and fixed) an issue where exported ASS files were missing procedural nodes. SItoA exports hair data in chunks (one chunk for every 200K hairs), but the exported ASS had just one procedural for the first chunk (chunk 0), but there should be one procedural for each chunk.

Here’s what the procedural for chunk 0 on frame 5 looks like. The missing procedurals would be for bin files like Hair.chunk.1.5.bin, Hair.chunk.2.5.bin, and so on.

procedural
{
 name Hair.SItoA.5000
 dso "sitoa_curves_proc.dll"
 data "//Projects/Support/Arnold_Scenes/Hair.chunk.0.5.bin"
 load_at_init on
}

So, for my first dive into the Arnold API, I put together a basic little script to add the missing procedurals. To do this, I had to learn how to:

  • Read and write ASS files
  • Iterate over nodes and find a specific type of node
  • Get parameter values from a node
  • Create new nodes
from arnold import *
import glob


ass_file = "Hair_Archive.ass"

AiBegin()
AiMsgSetConsoleFlags(AI_LOG_ALL)
AiASSLoad(ass_file, AI_NODE_ALL)

# Iterate over all shape nodes, which includes procedural nodes
iter = AiUniverseGetNodeIterator(AI_NODE_SHAPE);
while not AiNodeIteratorFinished(iter):
	node = AiNodeIteratorGetNext(iter)
	#print AiNodeGetName( node )

	# Is the node a procedural?	
	if AiNodeIs( node, "procedural" ):
	
		data = AiNodeGetStr( node, "data" )
		name = AiNodeGetStr( node, "name" )
	
		# Find all other chunk.<chunk-number>.<frame>.bin files
		chunks = glob.glob( data.replace( 'chunk.0', 'chunk.*' ) )

		# Add procedural nodes for chunks 1,2,3...
		for i in range(1,len(chunks)):
			n = AiNode("procedural");
			AiNodeSetStr(n, "name", "%s.%s" % (name, i) )
			AiNodeSetStr(n, "dso", "sitoa_curves_proc.dll")
			AiNodeSetStr(n, "data", data.replace( 'chunk.0', 'chunk.%s' % i ) )
			AiNodeSetBool(n, "load_at_init", True)

AiNodeIteratorDestroy(iter)


AiASSWrite(ass_file, AI_NODE_ALL, False)
AiEnd()

Exporting ASS files from Softimage with xsibatch


You cannot use xsibatch -export to export ASS files (because of the way sitoa implements ASS exporting). But it’s not too hard to do the same thing with SITOA_ExportScene and xsibatch -script.

This is what the xsibatch command line would look like (on Windows). For readability, I used the EXPORT_SCRIPT and SCENE variables to reduce the length of the xsibatch command line.

set EXPORT_SCRIPT = batch_export_scene.pys
set SCENE = \\server\project\scenes\elephant_herd.scn
xsibatch -processing -script %EXPORT_SCRIPT% -args -start 5 -end 20 -step 1 -scene %SCENE%

And here’s the batch export script.

I expose the basic parameters only, and I use the ASS Archive output path from the Arnold Render options, the scene name, and the [Frame] token to compose the output file name.

def main( start, end, step, scene ):
	Application.OpenScene(scene, "", "")
	x = Application.Dictionary.GetObject( "Passes.Arnold_Render_Options" )
	dir = XSIUtils.ResolveTokenString( x.output_file_tagdir_ass.Value, 0, False, None, None )

	scn = Application.ActiveProject.ActiveScene.Parameters("Name").Value	
	output = XSIUtils.BuildPath( dir, '%s.[Frame].ass' % scn )

	Application.SITOA_ExportScene( start, end, step, False, False, output )