View Full Version : Blender to RenderMan - Gladiola 'test'
06 June 2009, 05:32 AM
This started in the Blender section but since it has become more of an official WIP I decided to move the rest here.
Here is the original thread : http://forums.cgsociety.org/showthread.php?f=91&t=764325
Anyways. This is a collab project being done to test just how one would use Blender to model and animate a production, while using Aqsis to render the final frames.
So far it has been about a month into this and we have gotten quite a bit done. Today our pal Cedric dropped on us the spider rig for example, something that we have been anticipating for some time now. It is a beautiful rig as well, something I can animate myself and I for one am not the best at character animation.
One of my friends, who is a local low budget film maker had made the storyboards, which I imported into Windows Movie Maker to build a video storyboard so that my other friend in Wisconsin can do sound design, which in turn will allow us to begin syncing up animation to it. To get an idea of what the "story" is here is a first draft of vid storyboard which already is being edited, mainly timing for dramatic purposes but the overall feel remains the same...
Since the export script Mosaic is undergoing an upgrade, we are putting off some of the major work that needs to get done like texturing until that is given to us. Which really kind of puts us behind but that also lets us get some R+D work done, shaders built and I am also helping the Animux devs test out their Linux distro, which will ultimately save us in the end because they are letting us use their 12+ node renderfarm to do final frames on. So while some parts of this are not being worked on like we would want, there is still plenty of other stuff to do.
Next week I should be posting some major updates.
06 June 2009, 08:16 AM
I look forward to following your progress over the next few months. Your story is very cool, and you've got a great core team. Both Aqsis and Mosaic will be very battle tested by the end of this project, which is exactly what is needed for these apps.
I would love to contribute to the project, but with my available time, there's just no way it could happen. So I'll just appoint myself as the team cheerleader. I'll buy pom-poms.
The work you guys are doing is important for open source renderman to succeed. It's long overdue, and dammit, it looks like loads of fun!
06 June 2009, 12:25 AM
We still need people though, as it is our team consists of more devs than artists, though we are lucky enough to have the core people involved in the project. We are sorely needing an animator, as I myself am not good by any means, at least in the character animation dept. Physical and cloth sims oh yes I can do that lol.
I will update this thread very shortly.
07 July 2009, 11:15 PM
Well so far over the past few weeks a lot has been going on, despite the lack of "reporting" here.
I have been working with the Animux developers on their renderfarm, a 20 node powerhouse. We spent 2 weeks working on configuring DrQueue to run with Aqsis right, something that needed to get done now as opposed to later. We had a LOT of issues with it, however once we figured out what the heck was needed to get the process going, the lead programmer had written a simple shell script to "process" the batch script Mosaic creates. Another reason I had worked hard to get this renderfarm going was because I will be transfering assets to the Animux network and using the renderfarm now to perform the task of WIP renders (thus allowing me to not worry about tying up the CPU just to see what scenes will look like rendered in Aqsis). I am also switching to Animux for this production, Windows is just giving me problems and having at least a similar file structure in both my workstation and the remote renderfarm will keep things nice and tidy. Not to mention being able to use Cinepaint for some texturing (32bit displacements are important, GIMP does not support it).
Over the past week I have also been developing the first set layout. Originally the idea was to have two seperate train tracks, four tunnel entrances and a huge layout. I decided to try something smaller, as the set would just eat up resources, both in Blender as well in the render process. As you can see from the preview render below, there are no custom RSL shaders applied yet (waste of CPU cycles at this time), just the basic Mosaic surface shader that mimics Blender materials. The nice thing about this set is that most of everything in here is instanced, in Blender terms "Dupliaverted"... this cut down the time of exporting the objects more than half, as it only exports the one copy. As you can also see there is MUCH to be done. I expect that by next week this set will be completed. Already this image is obsolete and it was done last night!
So far the progress is going slow, though with the construction of this set I can start laying out some shots and seperate different scene files with x amount of objects in it (another reason Dupliverts are great - just delete vertices to get rid of object instances!)... so within a couple of weeks the shots can be finalized and can begin the process of animation.
Aside from the actual models, I have also been working on RSL shaders via Shaderman (since I am horrible at programming)
Mega Concrete Shader - still needs work
Rusty Surface - This shader is near complete
Rail - This complex shader works WELL on track rails or anything else requiring a smooth shiny top surface and dirty rusty bottom (or whatever one might find a use for)
And we continue on. A little behind schedule for what I was planning but a little closer to actually getting this done. I really doubt that it will die, even if I am the ONLY one on the artistic end doing this, it just might take longer than originally planned (we were shooting for end of August, though it might be end of Sept maybe even later).
07 July 2009, 12:30 PM
Very interesting project. I give you all my encouragement.
07 July 2009, 04:55 AM
Well it has been a long time since I have updated anything in here, again, due to real life situations on all the teams parts we haven't had a lot of time to keep the steam going BUT there has been progress!
One of the coolest things that has popped up in our pipeline is the Train. This beauty of a model has blown my mind. This model has been worked on over the course of two weeks entirely by Daniel, something I consider worthy of big budget film.
Now this is only the first version of the completed model, it had yet to be rigged or fully textured like we wanted. Well Daniel seemed to have read my mind, as I never mentioned anything and had redone the entire texture map. The result is this.
Now mind you these are Blender renders. We have been having some issues with our pipeline not finding texture maps, something I am correcting this week. There is also going to be a transfer to a full SVN server instead of the current Dropbox service that we have been using so far. In order to get the required data onto the Animux renderfarm it would be highly advisable to have this kind of control system, this way any time I would need to do test animations I can hop onto the network via VNC and use DrQueue to quickly produce previews. This will also be required when producing final frames.
Also due to my involvement with Animux, some of the 'Project Widow' assets are going to be showcased during the Animux Birds of a Feather meeting at SIGGRAPH 2009. Mark has asked me to produce some content for him and I felt it would be more powerfull to show our work than a bunch of tech demos with Suzanne. This will be a tense 2 weeks to get some content done in time but considering that modeling is almost completely done I think we can manage to produce at least 10 seconds of animation, not to mention screencasts of how to work with a 'real world' Blender to RenderMan production. Who can't argue with free publicity?? :)
Well till next time!
12 December 2009, 02:20 AM
It has been a LONG time since this has been touched here on CGTalk - mainly because over the past few months our work had ground to a stand still due to personal obligations and other things. However a lot of things have been done since the last post in July, so much so that it would be almost impossible to cover it all in one post so I will stretch it out over a series (over time of course).
First thing I wanted to address is that because of 'Project Widow' and our involvement with the Aqsis team, the recent Aqsis release actually has an updated OpenEXR display driver that handles multiple layers. This was a direct result of our desire to have it added, in fact during the summer and up until the release date we were testing out the daily builds. An improved Piqsl framebuffer was also developed because of our suggestions.
Mosaic was also worked on quite a bit during this time. Since the OpenEXR display driver was being worked on, Eric had also developed the handling of that particular driver into Mosaic itself so that there was no longer a need to manually type in exactly the kind of image type needed for a particular AOV function, instead it was all there ready to added to the scene export option as a preset.
As shown below in the Blender composite nodes - an example of the Multilayer OpenEXR display driver and how it was designed to work within Blender in post.
The thing that did fail, something that is still in the air, is the file control system. At first, Paul Gregory (creator of Aqsis) had developed a script based version control system called Arachnid, which was built on top of Unison. The testing did well at first with small files, though when it came down to working with the hundreds of files that have accumulated so far, some of them quite large in size, it started to choke. Currently we are still using the Dropbox application and that has served us very well simply because it is the only one that has not failed at all. SVN was also considered and still may be but so far Dropbox has performed well and on the free account we still have yet to fill our space to %50.
Now to the good stuff....
These images were from this past late summer, made to showcase during the Animux Bird of a Feather meeting during the 2009 SIGGRAPH.
The set design is pretty much the same, though some of the shaders have been reworked due to some issues, or visual appeal. Also - pay no mind to the render error in the second image, nor the lack of shadow maps in the last two images. The only REAL quality work was that first image in the series.
One of the newest ideas we are working on right now is using image maps in Blender to control varying amounts of dirt, grime or rust on several of the objects, mainly for visual appeal as some the shaders seemed too procedural and did not follow reality. This method also is allowing us to kind of "paint" the dirt on where we see fit, all of which is done by modifying the RSL code a bit. Grant you this is still in the "R+D" stage but so far our testing has gone fairly well.
Our main subject - the spider, really has NOT gone under any modifications since this past summer. The only thing worked recently is fixing up particular bad weight paints that control the hair. This image below is a render, just not with hair.
Mind you all these renders are done using Aqsis 1.6.0 (or a developmental version), none of them came from Blender itself.
Again - long overdue post and hopefully more to come very soon. Now that we are pretty much done with modeling, we are setting up scenes and will soon be working on animation. While our original timeline was for a few months of work, it looks as if this will go on well into next year, we are hoping that this will be complete by SIGGRAPH 2010. Time will tell though!
12 December 2010, 08:19 PM
It has nearly been a year since the last post, a lot has happened since then and I figured I would drop a line and give some updates about what is going on with this short film.
Firstly.... the title of "Project Widow" is officially called "The 10:15", why the name? Well the name that was given the spider was already in use by several films worldwide so that was pretty much a no go there. At first it was going to be "3:15" but again... another movie already had that name. Today was the projected release date of the film too but sadly this has not happened. So today is the day that the title itself is revealed.
A lot of programming went into developing the pipeline into what it is now. We are fortunate enough to have Shotgun Software hosting a server for us, they want to see this short film completed and have allowed us to use their service for free... how cool is that? Since we have begun to use the system tools have been created for Shotgun to use in our pipeline. Including a SVN post commit hook that sends info to Shotgun so we know exactly what file is modified and when, without having to have an email sent (though that function is sent to a few people for backup purposes). Also in the works is a Version thumbnail upload script that works directly in Blender, the first code was very simple and requires hand editing but it works fine.
The next version of the script has a GUI already and will be able to edit the Shotgun version id# and the change the thumbnail without having to hand edit the code.
We are also in the process of upgrading the pipeline from Blender 2.49 to 2.55. We are also taking over the development of RIBMosaic, Eric Back (Whiterabbit) has stopped working on the add on, which has now become a part of Aqsis. Us on the Widow project are developing the add on so that we can finish the short film.
Anyways.... another long awaited update done. We are more than halfway done now. It should not be much longer till the short is finished.
12 December 2010, 08:19 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.