Rafal Posted July 29, 2002 Share Posted July 29, 2002 Could someone explain me key difference between mantra and mantra3. Why does sidefx built-in 2 renderers? How about new format PIC supported by mantra3? Its not compatible with older one. Why? Quote Link to comment Share on other sites More sharing options...
stevenong Posted July 29, 2002 Share Posted July 29, 2002 mantra3 was, and still is, Houdni's renderer from version 1.0 while mantra was introduced together with Houdini version 4.0 to replace mantra3. I guess it's still bundled with Houdini for compatibility, in case a user needs to render something from before Houdini 4.0. The key difference? I would say differences. One of them is mantra allows displacement mapping while mantra3 only does bump mapping. The ability to use shaders is another. Also, you can pass attributes from Houdini to mantra so as to control shading, displacement etc. One of the biggest differences, IMHO, is the quality & speed. Higher quality for the same speed or same quality but much faster. You take your pick. I guess you mean the new PIC format output by mantra but not mantra3. Well, the new PIC format supports deep rasters which only mantra outputs. What are deep rasters? Here's a quote from the mailing list: You could also write your shader to take advantage of deep rasters for your composite, writing out the diffuse and specular (and other things too) to extra channels in your image. It doesn't seem to add that much time to a render (at least in any of the tests I've run on deep rasters). I guess deep rasters allow you to embed extra information which you can then use in Houdini COPs V2. Anyone? If I'm wrong on anything I said, please correct me. Thanks! Quote Link to comment Share on other sites More sharing options...
Marc Posted July 29, 2002 Share Posted July 29, 2002 Deep rasters will do exactly as you say. You can write out anything (and I mean anything) to the image as a seperate layer: lighting info, reflection, shadow, depth, normals or any arbitrary attribute. You can then bring this into Halo and use any of the composite operations to affect any of these layers. Its really something to behold... I'm hoping that some compositing videos will come out soon to show exactly how it all works. Marc Quote Link to comment Share on other sites More sharing options...
Rafal Posted July 29, 2002 Author Share Posted July 29, 2002 So when I rendered new PIC images. Whitch command you prefer to convert to usual format such TIFF, TGA, IFF. If you said I'd put into Halo this means there isn't anyone compositing system to handle this pics. You has to use HALO. Quote Link to comment Share on other sites More sharing options...
stevenong Posted July 29, 2002 Share Posted July 29, 2002 So when I rendered new PIC images. Whitch command you prefer to convert to usual format such TIFF, TGA, IFF. You can render straight out to TIFFs, TGAs or any format that Houdini supports by putting the extension (of the format) to the filename. Eg, mypic.$F3.tga will render out TGA files with a padded filename like such: mypic.001.tga, mypic.002.tga and so on. There is no need to convert. However, these formats will not support deep rasters. If you said I'd put into Halo this means there isn't anyone compositing system to handle this pics. You has to use HALO. I guess it all depends on whether other vendors add support for the PIC format so as to have access to the deep rasters. Else, you'll have to use Halo or COPs V2. Quote Link to comment Share on other sites More sharing options...
TheUsualAlex Posted July 29, 2002 Share Posted July 29, 2002 According to a few of Jason's Nuke listing, Nuke has Deep Raster support. So I gather that Nuke possibily support Houdini's native PIC format?? (That's just my guessing) Quote Link to comment Share on other sites More sharing options...
stevenong Posted July 29, 2002 Share Posted July 29, 2002 I think it's best to leave it to Jason, or anyone from DD, to answer the question though it's highly likely. Quote Link to comment Share on other sites More sharing options...
Marc Posted July 29, 2002 Share Posted July 29, 2002 I'm pretty sure it won't handle the new pic format, since its very new. Also DD has got their own fancy picture format called RGBEA (or something like that). And I believe that one is the 32bit deep raster one. Quote Link to comment Share on other sites More sharing options...
Jason Posted July 29, 2002 Share Posted July 29, 2002 NUKE will support the new deep .pic format as soon as Side Effects release the format specification - which should be happening in the next couple of weeks. I am a big fan of this format - it's very powerful. Our RGBEA format is not deep raster - it's an extension of the RGB format to support a common exponent so that we can render HDR images. It's based on Greg Wards RGBE format (from "Radiance") but supports alpha channel too. We currently spit out to multiple RGBEA format files as auxillary files when we want to render other shading information. Quote Link to comment Share on other sites More sharing options...
miguel m Posted July 30, 2002 Share Posted July 30, 2002 Just gessing here Jason, but that E channel stores some kind of "multiplier" that is applied to the other RGB? they are all 8 bit channel right? What format is a deep raster? an arbitrary numbrer of 8bit channels? maybe 32 per channel? Quote Link to comment Share on other sites More sharing options...
Jason Posted July 30, 2002 Share Posted July 30, 2002 Thats correct Miguel - the E channel is the 8-bit Exponent and the RGB are all 8-bit mantissa's. When the image is encoded, a suitable common exponent is found. This does introduce some quantizing if the magnitude of all three colours are vastly different, but it's acceptable for most things. Sometimes it doesnt work well and thats when we fall back on floating-point tiff. It's in line with the SGI format spec to have an arbitrary number of channels. The images will open up just fine in a package like photoshop which'll import all 5 channels. Of course the RGB and E channels look like garbage but the Alpha channel (the 5th channel) looks fine. We merely extended the RGBE format from Radiance to do this: RGBE format from Radiance Deep rasters mean that the format can support having multiple planes of arbtrary format data. e.g. Many RGBA layers, or an RGB layer, a Z-depth layer, a vector normal layer, etc. The .pic format can do this.. Quote Link to comment Share on other sites More sharing options...
miguel m Posted July 30, 2002 Share Posted July 30, 2002 Thanks Jason, RGBE seems very storage space friendly. Looks like i got to catch un with all these "superwhite" capable image formats... Quote Link to comment Share on other sites More sharing options...
Jason Posted July 30, 2002 Share Posted July 30, 2002 The .pic and .rat formats are capable of true floating point too, except it's not quantised and also capabile of negative numbers. The only real motivation for RGBE is storage concerns - 40bits of 8bit channels which compress easily with RLE compression. We maybe are considering moving to a true floating point format. For 95% of the cases RGBEA is fine, but falls down horribly when trying to export technical data like pixel-perfect depth passes and vector fields. And let me tell you, once you start to render and light with HDR, you'll never look back. J Quote Link to comment Share on other sites More sharing options...
miguel m Posted July 30, 2002 Share Posted July 30, 2002 I guess it may be much like a photographer does, setting HDR textured geometry around to emit light, not just the now universally used hemisphere thing... Quote Link to comment Share on other sites More sharing options...
Jason Posted July 30, 2002 Share Posted July 30, 2002 exactly... and the freedom at the compositing stage is amazing. Quote Link to comment Share on other sites More sharing options...
TheUsualAlex Posted July 30, 2002 Share Posted July 30, 2002 Isn't it that the general commercial monitors cannot display colors higher than 8 bit/channel? How does the compositing department deal with such? Although I am very much interested, I still haven't quite understand how to utilize image with arbitrary planes and floating points and HDR and such... Anyone care do talk about it a little bit more? Quote Link to comment Share on other sites More sharing options...
miguel m Posted July 30, 2002 Share Posted July 30, 2002 I believe that to view all the range of the image you just need to multiply it by a number smaller than 1, without cliping the end result of course. Jason, for an RGBE image only the E channel would need to be multiplied right? Quote Link to comment Share on other sites More sharing options...
Jason Posted July 30, 2002 Share Posted July 30, 2002 Thats right., the E is the exponent.. R = pow( R, E ) G = pow( G, E ) B = pow( B, E ) etc.. If you jsut scale down your images with a scalar less than one, you'd better make sure you use higher channel resolution to avoid quanisation. Like 16bit tiff or SGI16's. If you don't you'll compress away all your colour resolution. Quote Link to comment Share on other sites More sharing options...
MG Posted August 1, 2002 Share Posted August 1, 2002 I am starting to really love these forums! I was wanting to ask this (the difference between mantra and mantra3). But apparantly someone allready did it for me. Isn't that great! Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.