First, I was responding to someone who claimed that modelling and animating for shrek were done on Linux. So I wasn't bloody talking about off-line renderinging now, was I.
3D accelerators of the sort found in most PCs would not help with off line rendering for the most part. However, the GSCube does not work like a normal PC video card. Instead of having a dedicated rendering chip, it has many CPUs that are based the on the Mips R5000 CPU, but with additional vector units. Now, historically, raytracing has not been something that vectorizes very well. Scan line rendering does. So, while the frame buffers of a GSCube would be wasted, the stuff that makes a GSCube good for video games would also make it good for rendering movies, as long as scan line rendering is still a popular way to render movies.
Re:Books about Graphic Software...
on
The Blender Book
·
· Score: 1
Faster than what? Faster than Truespace? I agree. Faster than Softimage 3D? I don't think so.
In the context used, faster refered to how fast one can work, not how fast the program runs (although that is also important). It didn't take me anytime to become productive in SI 3D (never touched XSI), but I still stumble slowly through blender. I'm not sure what the difference is. It just doesn't quite feel right. But, at that price I will persevere.
First http://www.geocities.com/SiliconValley/Server/9029/images4.html is a link to a very large gallery of images from the movie. There is even one of the Aki naked, although I am not sure why it got made in the first place. http://www.geocities.com/SiliconValley/Server/9029/aki_nude.jpg is the direct link.
Well, it is fairly standard practice to model humanoid characters nude. Perhaps the details aren't needed, but it doesn't seem that uncommon for them to be put in anyway. That is why a nude model exists. And well, once it exists, you know people are likely to get carried away.
I am surprised there were no articles on Shrek and Linux because it was done not only using linux for the rendering but for some of the actual animating, modeling, etc. as well
What hardware did they run it on? It seems to me that while linux is extremely useful (I use it alot at home) that it's 3D support still isn't up to snuff. At least, not from what I've seen of Matrox, ATI, tdfx, and nvidea. For instance, my understanding is that geometry acceleration just plain isn't supported for any card under XFree yet.
Now, I've heard rumors that HP's linux machines, the ones shipped with the FX cards, use HP-UXs X server instead of Xfree. I'd be interested in seeing that confirmed somewhere.
Re:Blender book... no PDF, no GPL
on
The Blender Book
·
· Score: 1
From time to time, I work on writing my own animation software. But, I don't want to have to also write my own modeller, and none of the libre ones I've seen are really that good. So, I'm trying to learn blender well enough to do modelling in it.
Re:Books about Graphic Software...
on
The Blender Book
·
· Score: 2
When people say that, I want to know what they are comparing against. For instance, is blender faster than Truespace? What about Lightwave 3D? 3DS Max? All quite possibly. But, it is it faster to use than SoftImage 3D or XSI? Well, I haven't used XSI yet, but I doubt that it is faster than SI 3D. How about is it faster than Maya? Mirai? I doubt those two, but I haven't ever had a chance to try them. The s-mesh features though are a good sign. I far prefer sub division surfaces to NURBS.
I remeber when Lightwave came out with MetaNURBS. Many people were bashing it because it wasn't really NURBS, but was instead subdivision based. Now, everyone has realized the problems with NURBS (complexity issues) and is racing to implement Sub-division modelling.
I thought the idea was to teach people how to use computers, not how to blindly follow some steps that lead to results on only certain configurations. Back when I still ran Windows, some people would come to my house, sit down at my computer, and be parallelized because it didn't look exactly the way they expected. I didn't do anything wierd, other than rearrange the templates in office, and completely rearrange the start menu. Office had numerous changes made, but only of the kind that added new things, not stripping away old menu items.
I'm waiting to be able to find a rack on sale where I live. When I look at my gear, I see the following things that will go in the rack. First, on the bottom (sitting on the floor in the rack) is a Sun 3/160. This thing is a monster. One top of that will be a rack mountable APC UPS. On top of that a 4u FDDI concentrator, on top of that a 1 or 2u ethernet switch with fddi uplink (I haven't purchased this piece yet). Then, there will be a Sparc 2, and 1 file server, and one workstation, and some misc gear (serial switches, DSL modem, gateway device, etc), and if there is room, a dumb terminal. The whole thing would be a bit more reasonable if I weren't trying to place a Sun 3/160 in the bottom. If it turns out to be a problem, I guess I can move it to the other end of the desk. But it really is a nifty machine. It has a 68020 CPU. I want to find some of the cooler accessory boards for it, like the floating point accelerator, and the graphics buffer and processor cards. Maybe even a Symbolics UX400 board.
And I know people with far more rack gear than I own. These people usually have a home machine room though.
Actually, sometimes configure compile install run doesn't work so well. I've spent the past week trying to install enough stuff to get Mozilla, GnuCash, and Gnumeric to run. I've had to tweak the source code to get some things to compile (libgpopt for one), and an currently stuck on GAL (which seems to be complaining about berkeley db problems, but I haven't been able to figure out exactly what is the problem).
The biggest issue is the lack of clear direction of what order to install things in. So, I know that Gnumeric requires glib and gtk, so I install those, then try building Gnumeric. So, it complains about a lack of libole2, so I get that and it complains about missing 10 things. I fix those, and now Gnumeric complains about missing another library, and so on. I'm hoping that GAL is the last one.
I'm using Debian testing by the way. I just prefer a source installed gnome since than I know exactly what versions of what packages and where they are installed.
That looks so cool. I just wish that I could find such innovation in desktop machines. I would love to be able to get one of those boxes with Riva TNT or better graphics.
A friend works for a local printing company (the one that does Readers Digest), and I got the impression from him that they used mostly custom written software since off the self stuff didn't do things they needed, like sort the print order by street address so that packs of material could be bundled in such a way so that the mailman just strolls down the street taking the top catalog off the stack and always finding that it is the right one for the next house.
I think my biggest problem with Quake 2 was that the controls never quite felt right. Now, the techno level designs was fine, but I also had some minor problems with the renderer. I think this is more of a tast issue than anything, but the renderer looked to smooth in some ways.
Avid style editing was always on the desktop. That was the whole point of Desktops.
You probably already did this, but if not, get the directors commentary DVD of El Mariachi. It is extremely fascinating. I usually watch more than a few scenes of a director's commentary, but this was so jam packed that I watched the entire films commentary.
To save money, Robert Rodriguez would use watch ever he could of a take. So, if I ran down a ramp and threw a guitar at a balcony and missed, well he decides that is where a cut in the final edit will be. So he moves the camera and the has someone off screen through the guitar case over, and he cuts between the two just before it was obvious that my throwing the guitar case would have missed.
Also, Robert Rodriguez never made a film print of El Mariachi. He had it transfered to video and edited it there. He then sent the video around with the idea that whoever bought it could pay to have an actual film edit done, since getting a print made would have cost nearly as much as the entire movie itself.
Finally, on a seperate note, while Spy Kids was a bit to cheerful and child oriented at times, it still was a rather amusing movie, and time reasonably well spent. I'm sure that all the parents who took their kids were extremely happy that for once it was something they could rather enjoy also.
Luxo Jr. was sent to 16mm film, not 35. Further, considering the age (and cheapness), it wouldn't surprise me if it was only 10bit linear. Still that's 4 times better than 8bits. Frankly, I'm now pretty sick of 8bit graphics too much of the time. But I don't have to money for an Onyx or Octane, and that is pretty much the only way to upgrade past 8bits.
>Raytracing isn't even the best you can do as it
>can't cater for atmospehric effects and
>diffusion of light through a scene.
I thought that raytracing was supposed to be the best way to deal with these issues.
For instance, for atmosperic effects, you can deal with this two ways. You can have the atmosphere be defined as a set up points(or really small polygons) that is just a little spec that is mostly translucent, but that reacts to being hit by light. This is perhaps the most acurate way of dealing with atmospheric effects. The other way is to have an atmospheric effect defined as a large solid object that is essentially a vector field for the direction and color and intensity vectors of the light rays that pass through it.
The other thing to help improve the quality of raytracers is to make them actually calculate and trace the diffusion of rays. This is somewhat related to radiosity, but I'm not exactly sure how. I'm not an expert on radiosity, but I've read a few documents on the subject. Unfortunatly, these documents mainly had to do with static radiosity (the radiosity is calculated once and stored as light values attached to each vertice) for real time rendered stuff.
For years, I ignored rendering technology hoping that the next version of POVray would do what I wanted. Well, it won't, so I think it is about time for someone to actually commit to writing a shader based (perhaps Renderman compatible) rendering engine. Lots of people start and make a lot of noise, but then they drop it. Maybe one day I will start a project and actually finish it. Getting a basic raytracer going can't be hard. It's writting the shading engine and difussion engines that would be rather difficult.
I don't work at Pixar, so I can't say what they actually do, but I do know their software somewhat (used to be bundled with Nextstep before Apple bought Next). Their software supports defining everything in a scene once and using one file for the whole scene, than moving the objects in the scene. For meshs that change (think all the characters), I don't know if there is a good way to define their mesh once, or if you need their mesh redefine in each frame (which is to say, I don't remeber if it is possible to have a Shader move the vertices for skeltal animation). However, that doesn't mean that Pixar doesn't use one file for one frame. It just means that their software doesn't require it.
Pixar doesn't raytrace. They are very clear about the fact that their renderer is a scanline renderer. I don't know if they use radiosity or not, but whether or not they do doesn't say anything about whether they ray trace, since from what I've seen, much radiosity is done as a preprocessing step and stored in the light value fields of each vertice. This means that if nothing in a room changes position, you can keep rerendering a moving camera without repeating the radiosity step. This is usefull for more realistic lighting in games. Prerender as much radiosity as possible, then just use cheap tricks for shadows and lighting for the moving characters.
Actually, Halflife was mainly written in the Quake 1 engine with some bits and pieces later borrowed from Quake2. When Halflife was started, the Quake 2 engine wasn't available.
Personally, I prefer Quake 1 to Quake 2. But Quake 3 is stinkin cool.
Tell you friends to look into lisp and smalltalk. Lisp has had a lot of powerful graphics and interactive applications written in it. But, currently the free lisps lack a comprehensive environment of content tools. However, Allegro CL from Franz (www.franz.com) should meet your friends needs if they can afford it. And to people who say lisp is outdated, well look at how AllegroCL is being used in the case studies at Franz's web site.
And if functional programming languages with Lots of Irritatingly Silly Parentheses isn't there thing, have the try Smalltalk, specifically, Squeak (http://www.squeak.org/). Among places that use squeak is Disney. Squeak has many authoring features, plus on Unix systems it is easy to use extra devices (since virtually all devices are files, and any programming language can use files, even if it knows nothing about the device the file represents).
Actually, everyone who wants to program should learn one or both tools fairly well. I am convinced that the path to being a great program lies in understanding the machine at a low level (IE, understanding how to do stuff in ASM), and in learning the great programming languages of history. Further, both lisp and smalltalk lend themselves to spike solutions which is important in maintaining the mental state called flow. Which is to say, they support creative thinking. Once familiar with the environment, you can just do things, and then if they work, refactor them into proper structures.
Re:Stop picking on poor Angelina's chest
on
Review: Tomb Raider
·
· Score: 1
Never seen the stand. Personally, I thought that they should have picked Rhona Mitre for the part of Lara though. I've seen some movies with here and she is a decent actor, and why she played Lara at conventions I thought she was dead on.
But, after Tomb Raider 3 pissed me of for so many reasons that I no longer really care.
"The fact is that the American people are sick and tired of this whole foreign aid concept anyhow"
I fail to see what is neanderthalic about that? The fact is that at least THIS american person is rather sick of the whole foreign aid concept.
Why are we spending more to be less effective than civilian aid groups (thinking of many christian fundamentalist groups that seldom offer aid without missionaryish strings attached). If the recieving country doesn't want whatever propaganda the civilian aid group might have, well, then let them refuse it and do with out aid.
Beggers should be choosy about where their aid comes from.
Talk about inflexible. For all the software I could find on those pages, I would need to first re arrange all my drives (making linux non bootable until I put them back), and then erase my linux and other OS partitions just so that I could install windows to run these programs before reformatting again to replace linux.
Why don't more of the people practice their hacking under linux. We have superior tools for one thing, and their work would be more appreciated here is another thing.
I think that almost anyone who wants to spend a little time on it can become a decent sound engineer. While a natural sound engineer with perfect pitch may be head above this more common person, the common guy could still produce some pretty nice recordings.
I mean, look at many blues and jazz and Alternative CDs. They were produced on a far smaller budget and less professional facilities than Britany Speers is, yet who is the higher paid engineer.
I think that the only major place where there is no way around getting a highly trained and experienced and expensive professional is when recording classical music and opera. Those two rely on understanding the mechanics of large rooms which are a lot more complex than figuring out how to mic a guitar amp in a small theatre.
Slashdot Mirror
First, I was responding to someone who claimed that modelling and animating for shrek were done on Linux. So I wasn't bloody talking about off-line renderinging now, was I.
3D accelerators of the sort found in most PCs would not help with off line rendering for the most part. However, the GSCube does not work like a normal PC video card. Instead of having a dedicated rendering chip, it has many CPUs that are based the on the Mips R5000 CPU, but with additional vector units. Now, historically, raytracing has not been something that vectorizes very well. Scan line rendering does. So, while the frame buffers of a GSCube would be wasted, the stuff that makes a GSCube good for video games would also make it good for rendering movies, as long as scan line rendering is still a popular way to render movies.
Faster than what? Faster than Truespace? I agree. Faster than Softimage 3D? I don't think so.
In the context used, faster refered to how fast one can work, not how fast the program runs (although that is also important). It didn't take me anytime to become productive in SI 3D (never touched XSI), but I still stumble slowly through blender. I'm not sure what the difference is. It just doesn't quite feel right. But, at that price I will persevere.
Well, it is fairly standard practice to model humanoid characters nude. Perhaps the details aren't needed, but it doesn't seem that uncommon for them to be put in anyway. That is why a nude model exists. And well, once it exists, you know people are likely to get carried away.
What hardware did they run it on? It seems to me that while linux is extremely useful (I use it alot at home) that it's 3D support still isn't up to snuff. At least, not from what I've seen of Matrox, ATI, tdfx, and nvidea. For instance, my understanding is that geometry acceleration just plain isn't supported for any card under XFree yet.
Now, I've heard rumors that HP's linux machines, the ones shipped with the FX cards, use HP-UXs X server instead of Xfree. I'd be interested in seeing that confirmed somewhere.
From time to time, I work on writing my own animation software. But, I don't want to have to also write my own modeller, and none of the libre ones I've seen are really that good. So, I'm trying to learn blender well enough to do modelling in it.
When people say that, I want to know what they are comparing against. For instance, is blender faster than Truespace? What about Lightwave 3D? 3DS Max? All quite possibly. But, it is it faster to use than SoftImage 3D or XSI? Well, I haven't used XSI yet, but I doubt that it is faster than SI 3D. How about is it faster than Maya? Mirai? I doubt those two, but I haven't ever had a chance to try them. The s-mesh features though are a good sign. I far prefer sub division surfaces to NURBS.
I remeber when Lightwave came out with MetaNURBS. Many people were bashing it because it wasn't really NURBS, but was instead subdivision based. Now, everyone has realized the problems with NURBS (complexity issues) and is racing to implement Sub-division modelling.
I thought the idea was to teach people how to use computers, not how to blindly follow some steps that lead to results on only certain configurations. Back when I still ran Windows, some people would come to my house, sit down at my computer, and be parallelized because it didn't look exactly the way they expected. I didn't do anything wierd, other than rearrange the templates in office, and completely rearrange the start menu. Office had numerous changes made, but only of the kind that added new things, not stripping away old menu items.
I'm waiting to be able to find a rack on sale where I live. When I look at my gear, I see the following things that will go in the rack. First, on the bottom (sitting on the floor in the rack) is a Sun 3/160. This thing is a monster. One top of that will be a rack mountable APC UPS. On top of that a 4u FDDI concentrator, on top of that a 1 or 2u ethernet switch with fddi uplink (I haven't purchased this piece yet). Then, there will be a Sparc 2, and 1 file server, and one workstation, and some misc gear (serial switches, DSL modem, gateway device, etc), and if there is room, a dumb terminal. The whole thing would be a bit more reasonable if I weren't trying to place a Sun 3/160 in the bottom. If it turns out to be a problem, I guess I can move it to the other end of the desk. But it really is a nifty machine. It has a 68020 CPU. I want to find some of the cooler accessory boards for it, like the floating point accelerator, and the graphics buffer and processor cards. Maybe even a Symbolics UX400 board.
And I know people with far more rack gear than I own. These people usually have a home machine room though.
Actually, sometimes configure compile install run doesn't work so well. I've spent the past week trying to install enough stuff to get Mozilla, GnuCash, and Gnumeric to run. I've had to tweak the source code to get some things to compile (libgpopt for one), and an currently stuck on GAL (which seems to be complaining about berkeley db problems, but I haven't been able to figure out exactly what is the problem).
The biggest issue is the lack of clear direction of what order to install things in. So, I know that Gnumeric requires glib and gtk, so I install those, then try building Gnumeric. So, it complains about a lack of libole2, so I get that and it complains about missing 10 things. I fix those, and now Gnumeric complains about missing another library, and so on. I'm hoping that GAL is the last one.
I'm using Debian testing by the way. I just prefer a source installed gnome since than I know exactly what versions of what packages and where they are installed.
That looks so cool. I just wish that I could find such innovation in desktop machines. I would love to be able to get one of those boxes with Riva TNT or better graphics.
A friend works for a local printing company (the one that does Readers Digest), and I got the impression from him that they used mostly custom written software since off the self stuff didn't do things they needed, like sort the print order by street address so that packs of material could be bundled in such a way so that the mailman just strolls down the street taking the top catalog off the stack and always finding that it is the right one for the next house.
I think my biggest problem with Quake 2 was that the controls never quite felt right. Now, the techno level designs was fine, but I also had some minor problems with the renderer. I think this is more of a tast issue than anything, but the renderer looked to smooth in some ways.
Avid style editing was always on the desktop. That was the whole point of Desktops.
You probably already did this, but if not, get the directors commentary DVD of El Mariachi. It is extremely fascinating. I usually watch more than a few scenes of a director's commentary, but this was so jam packed that I watched the entire films commentary.
To save money, Robert Rodriguez would use watch ever he could of a take. So, if I ran down a ramp and threw a guitar at a balcony and missed, well he decides that is where a cut in the final edit will be. So he moves the camera and the has someone off screen through the guitar case over, and he cuts between the two just before it was obvious that my throwing the guitar case would have missed.
Also, Robert Rodriguez never made a film print of El Mariachi. He had it transfered to video and edited it there. He then sent the video around with the idea that whoever bought it could pay to have an actual film edit done, since getting a print made would have cost nearly as much as the entire movie itself.
Finally, on a seperate note, while Spy Kids was a bit to cheerful and child oriented at times, it still was a rather amusing movie, and time reasonably well spent. I'm sure that all the parents who took their kids were extremely happy that for once it was something they could rather enjoy also.
Luxo Jr. was sent to 16mm film, not 35. Further, considering the age (and cheapness), it wouldn't surprise me if it was only 10bit linear. Still that's 4 times better than 8bits. Frankly, I'm now pretty sick of 8bit graphics too much of the time. But I don't have to money for an Onyx or Octane, and that is pretty much the only way to upgrade past 8bits.
>Raytracing isn't even the best you can do as it
>can't cater for atmospehric effects and
>diffusion of light through a scene.
I thought that raytracing was supposed to be the best way to deal with these issues.
For instance, for atmosperic effects, you can deal with this two ways. You can have the atmosphere be defined as a set up points(or really small polygons) that is just a little spec that is mostly translucent, but that reacts to being hit by light. This is perhaps the most acurate way of dealing with atmospheric effects. The other way is to have an atmospheric effect defined as a large solid object that is essentially a vector field for the direction and color and intensity vectors of the light rays that pass through it.
The other thing to help improve the quality of raytracers is to make them actually calculate and trace the diffusion of rays. This is somewhat related to radiosity, but I'm not exactly sure how. I'm not an expert on radiosity, but I've read a few documents on the subject. Unfortunatly, these documents mainly had to do with static radiosity (the radiosity is calculated once and stored as light values attached to each vertice) for real time rendered stuff.
For years, I ignored rendering technology hoping that the next version of POVray would do what I wanted. Well, it won't, so I think it is about time for someone to actually commit to writing a shader based (perhaps Renderman compatible) rendering engine. Lots of people start and make a lot of noise, but then they drop it. Maybe one day I will start a project and actually finish it. Getting a basic raytracer going can't be hard. It's writting the shading engine and difussion engines that would be rather difficult.
I don't work at Pixar, so I can't say what they actually do, but I do know their software somewhat (used to be bundled with Nextstep before Apple bought Next). Their software supports defining everything in a scene once and using one file for the whole scene, than moving the objects in the scene. For meshs that change (think all the characters), I don't know if there is a good way to define their mesh once, or if you need their mesh redefine in each frame (which is to say, I don't remeber if it is possible to have a Shader move the vertices for skeltal animation). However, that doesn't mean that Pixar doesn't use one file for one frame. It just means that their software doesn't require it.
Pixar doesn't raytrace. They are very clear about the fact that their renderer is a scanline renderer. I don't know if they use radiosity or not, but whether or not they do doesn't say anything about whether they ray trace, since from what I've seen, much radiosity is done as a preprocessing step and stored in the light value fields of each vertice. This means that if nothing in a room changes position, you can keep rerendering a moving camera without repeating the radiosity step. This is usefull for more realistic lighting in games. Prerender as much radiosity as possible, then just use cheap tricks for shadows and lighting for the moving characters.
Actually, Halflife was mainly written in the Quake 1 engine with some bits and pieces later borrowed from Quake2. When Halflife was started, the Quake 2 engine wasn't available.
Personally, I prefer Quake 1 to Quake 2. But Quake 3 is stinkin cool.
Tell you friends to look into lisp and smalltalk. Lisp has had a lot of powerful graphics and interactive applications written in it. But, currently the free lisps lack a comprehensive environment of content tools. However, Allegro CL from Franz (www.franz.com) should meet your friends needs if they can afford it. And to people who say lisp is outdated, well look at how AllegroCL is being used in the case studies at Franz's web site.
And if functional programming languages with Lots of Irritatingly Silly Parentheses isn't there thing, have the try Smalltalk, specifically, Squeak (http://www.squeak.org/). Among places that use squeak is Disney. Squeak has many authoring features, plus on Unix systems it is easy to use extra devices (since virtually all devices are files, and any programming language can use files, even if it knows nothing about the device the file represents).
Actually, everyone who wants to program should learn one or both tools fairly well. I am convinced that the path to being a great program lies in understanding the machine at a low level (IE, understanding how to do stuff in ASM), and in learning the great programming languages of history. Further, both lisp and smalltalk lend themselves to spike solutions which is important in maintaining the mental state called flow. Which is to say, they support creative thinking. Once familiar with the environment, you can just do things, and then if they work, refactor them into proper structures.
Never seen the stand. Personally, I thought that they should have picked Rhona Mitre for the part of Lara though. I've seen some movies with here and she is a decent actor, and why she played Lara at conventions I thought she was dead on.
But, after Tomb Raider 3 pissed me of for so many reasons that I no longer really care.
Dude, just watch where she goes online. I mean, really watch, like being in the room, or at least popping your head in frequently.
That implies more trust than any other method does. At least you trust her enough to be honest with her this way.
"The fact is that the American people are sick and tired of this whole foreign aid concept anyhow"
I fail to see what is neanderthalic about that? The fact is that at least THIS american person is rather sick of the whole foreign aid concept. Why are we spending more to be less effective than civilian aid groups (thinking of many christian fundamentalist groups that seldom offer aid without missionaryish strings attached). If the recieving country doesn't want whatever propaganda the civilian aid group might have, well, then let them refuse it and do with out aid. Beggers should be choosy about where their aid comes from.
Talk about inflexible. For all the software I could find on those pages, I would need to first re arrange all my drives (making linux non bootable until I put them back), and then erase my linux and other OS partitions just so that I could install windows to run these programs before reformatting again to replace linux.
Why don't more of the people practice their hacking under linux. We have superior tools for one thing, and their work would be more appreciated here is another thing.
40 songs and 40 years. That's 40x40=1600, not 160.
I think that almost anyone who wants to spend a little time on it can become a decent sound engineer. While a natural sound engineer with perfect pitch may be head above this more common person, the common guy could still produce some pretty nice recordings.
I mean, look at many blues and jazz and Alternative CDs. They were produced on a far smaller budget and less professional facilities than Britany Speers is, yet who is the higher paid engineer.
I think that the only major place where there is no way around getting a highly trained and experienced and expensive professional is when recording classical music and opera. Those two rely on understanding the mechanics of large rooms which are a lot more complex than figuring out how to mic a guitar amp in a small theatre.