Friday, November 10, 2006

The Manufactured Image: Ideas for Change (part 2)

cont'd from part 1...

- Get lead modelers to create second pass rough Cg sets, but do this while being informed by the camera and pencil layout data. So this is another place for technology to come into play- develop a system of exporting rough camera and pencil/color layout info for import into any asset at a later time. Get this working so that camera data can begin to be stored and shared as early as layout or concept development for a set. If you’re smart you can save your story artist exploratory camera data as well. The sooner you can capture camera data for use downstream the better.

- Following pencil layout (concurrent with color tonal layout) put your modeling team to work on individual asset creation. Approve the individual assets in context based on looking at them from the various camera angles exported from previous steps. Hold off on materials and textures, though. We’ll get there in a moment.

- Master lighting for sets could be a two stage process. Stage one would be rough master lighting- ie: before textures and materials are applied to the objects in the modeled set. Yes, light the set without any textures at all. I used to do this a long time ago when I was doing Cg illustration because I found it easier to see the macro level art in the image without all the details cluttering things up. The idea is to shift away from a micro focus (textures, materials on individual assets) and onto a macro focus. Paint with light to build the environment. As with modeling, do this as informed by published rough camera data and in scene color key projection maps created by the colorists. Now you use broad strokes of color and tone to build your compositions. Materials and textures still wait.

* A word on the camera data to this point: It’s all rough. Yes, it will get tightened down & changed a bit later. Yes a fair amount of work will be thrown out and re-done. This is no different than the current system. But a rough idea of how something will look is better than no idea at all.

- After rough master lighting, NOW you can make informed material and texture callouts for the detail tasks. The individual texture artist now has not just concept drawings, but camera specific tonal, color and lighting cues to work from. They can better accomplish their tasks while working in concert with the larger macro artistic focus instead of getting lost down the rabbit hole of pretty details in isolation. Again, approvals from real camera views, not isolated turn table renders.

- Give supervisors provisional approval power to get assets out of the isolated setting and put into the set in context. Basically never allow an art director or director to look at a set dressing asset all by itself in a grey background render. This is a recipe for noodling. Final approvals should be in context only (as much as humanly possible). For approvals the art director and director can see how these things all play together and in the dailies theater by having the editor toggle from pencil layout projection pass to colorist color/tone projection pass to no materials rough master lighting pass to detailed asset pass -all in moments. All notes for revision to individual assets are given in light of the scene/camera view context. Look at the set with sub assets from as many camera views as you can for a set. (remembering that we already have rough master lighting here). I guarantee the results will be better. Producers will love me because this one thing alone will cut the revision budget for materials and textures by a third- at least.

- Second pass master lighting can be done to tighten the integration of details and macro art elements after all the assets for a set are textured & approved in the above step. All along this process the camera data is updated and made stronger via layout, animation, etc. As camera data gets significantly adjusted new rough pencil layout & colorist projections can be made if necessary. However somewhere along this stage of the game (or slightly before) the pencil stuff fades away and is no longer updated because the CG imagery is starting to take stronger shape.

- Using “same as” camera information, pre-render your background plates for scenes that do not require character/set interaction. Build a library of rendered backgrounds for scenes. Then, to better the compositon, use colorists to punch up the images digitally. Push areas darker, pull saturation more in other areas. You can even digitally paint out things that make the composition klunky. What would take hours to do in test renders and lighting adjustments can be done in minutes in a digital paint program. This is an example of mid-pipeline compositing that could work.

- To ease the crunch on scene lighting, get scene lighting happpening sooner. Don’t wait for all the textures and details to be done to begin scene specific lighting. Make scene lighting a two stage process like master lighting- first stage sans texture details, second pass after details are done. Do first pass rough scene lighting for each scene based on the rough master set lighting. Scene lighting can work in parallel with animation, immediately following Cg layout. Most studios have a technological system to allow scene lighting to occur in parallel to animation. Rather than building a tsunami of inventory for lighters, work down the task load in phases. Rough scene lighting (sans materials and textures) will again give scene lighters the ability to macro focus on the artistry of their scenes and use the effective “big brush” of light color to build better compositions.- Second pass scene lighting occurs after all assets for a scene are detailed and textured. But since the broadstroke lighting for the scene has been handled in rough scene lighting second pass scene lighting can go faster to tighten down the details and how things integrate, making life more humane for lighters at the end of the show.

- Create tools that allow scene lighters to macro adjust base colors for assets in a specific scene. Got a scene that needs everything to be a cool green to the left, but all the assets are textured with warmer colors? Or is your character feeling too much like the background and your lighting setup is only getting you so far in helping them stand out? Allow lighters the ability to marquee select a section of the screen (or specific objects) and hook up a master HSV modifier into all the materials & textures for those objects- including characters and props. To go one better make it have depth capabilities as well (ie: stronger adjustment power further back in z space, etc.). Go even one better-er, give it basic masking functions like feather, radial or linear gradients. Make a similar tool to macro control specular highlights or bump map depth or texture detail contrast & brightness. Basically allow the lighters to have more tools to “paint” the scene than just lights and give them access to make gross adjustments to object materials on a macro level. It’s not about them sending things back for fixes, nor is it about opening up every texture or material detail for a lighter to re-interpret. It’s about giving them simple, easy to use macro tools to quickly make compositional improvements right at their workstations without creating an avalanche of re-do orders. Make your lighters your army of colorists. Make the scene specific results publishable so they can be propagated across all same-as and similar-as scenes. Production managers can thank me later.

-Why not use late stage compositing to fix problems? You can, but compositing by nature tends to need many layers and elements in final-ish render form to be of best use. A good compositor can rescue almost anything. Problem is by the end of the pipeline they rarely have the time (or disk space for all the elements). Do rough compositing along the way? Now you’re fighting for hardware assets like network bandwidth and storage capacity as you render rough layers for many, many stages of the show. The data storage and tracking tasks become unwieldy. Compositing is used extensively for lower volume projects like commericals or VFX jobs. A commercial typically will have fewer than two dozen scenes total, so compositing makes economic sense. And a big VFX job is 250+ scenes. A monster VFX job is anything over 500 scenes. But a typical 95 minute Cg animated feature film is well north of 1,200 scenes. With each 2k film image taking up around 7mb of space on disk, every comp element significantly ups the needed size of your server space. The economies of scale work in reverse here. In other words, there is a reason why for years Pixar rendered everything “in frame”. Massive use of anything more than the most basic of scene compositing on a feature film scale is murder on your infrastructure budget.

Anyhow, those are some ideas. How many will work? I don’t know. I do know a number of them will work and are very do-able. These certainly aren’t the only ideas or solutions possible. They’re just the ones I have in my head or have tried. Several years back when I was installed as CG Supervisor for Big Idea’s second feature film project (which never got made, sadly- the studio went bankrupt before we could get very far into production) we tried many different things to get the films looking better and make the pipeline more artist friendly. Other ideas have come to me in the years since. I’ve listed most of them here in the hope that other bright people will use them as jumping off points for even better ideas. Others will no doubt find it easy to poke holes in the ideas as well. All the same, I’m just some dude who’s willing to stick my neck out and say the current system isn’t all that hot and here’s how I think it can be better. That’s a conceit that assumes a lot of responsibility, and I hope I’ve handled it well. And if I’m wrong, well… nobody can fault me for lack of trying.

Thanks to Mark Mayerson, Amid Amidi and all the others who have taken the time to link back to the post series over the last few weeks. Those of you new to my humble little corner, thanks for coming! I hope you found something useful in all of this rambling. Hopefully we can see the artistry in feature CG films pushed in ways that challenge technological paradigms. I’m eager to see things yet unseen.

1 comment:

Keith Lango said...

original comments here...