I got back in late into the 3D panel after taking care of some work on the show floor. Here’s my raw, unfiltered, typos-and-all notes from the panel on 3D, that I jumped in on late, so missed a lot of names. Some really good info, especially about the importance of modulating and babysitting the convergence point cut to cut, and how that affects editorial, and what steps to take to make sure the 3D cut works.
Howard Lutz –
3D – silver screens – manufacturing quality has affected display quality
supply of 3D camera rigs at the moment is an issue
-if there’s 2 major films being done in the Hollywood area, is OK, the minute a 3rd is in production, not enough cameras
-it has ramped up so fast, “Is it real or a fad?” It is real, there’s a commitment from CG and live action, it would be good to see a better supply of stereoscopic camera rigs, as well as production and post tools
-Dave from ??? on challenges of getting 3D from big to small screen
-whether dimensionalists, shooters, or renderers, we end up with a left and right view, superimposed, that cause your eyes to converge in front/at/behind the screen
-images that create binocular disparity, when done right is gorgeous, when done wrong is painful
-fundamental disparity – focusing on screen 40 feet away, converging elsewhere. Sends confusing depth cues to brain. Convergence wins…except on television
-and then, a formula on a slide, about left and right eye, separated by about 2 1/2 inches. Left eye crosses to right and vice versa to converge on the apparent distance to the image, which is different from the distance to the SCREEN
-Adult’s interocular distance – 2.5 inches, child is 2.0 inches
-if you have a face behind screen, and bring it closer, your eyes expect more difference and “roundness” of the thing’s info – is an issue on small screens
-Jim Maynard at Dreamworks is working on fixing this issue in software automatically in his renders
-it is the little d number – the apparent distance to object, as it compares to the big D, the distance to the actual screen
-5 feet from screen, closest object can be 89%, furthest at 114
-depth budget – how much can you go in front or behind of the screen – shown as a percentage of size of screen
-you can make your eyes diverge if you try to run it too much, big to small screen
-there’s a bell curve of acceptance to “Fundamental Disparity” – the eyes’ converging muscles and the eyes’ focusing muscles sendign conflicting depth into the brain. a few percent of folks have sensitive eyes – at the other end, agile eyes can handle anything
-if you look from dashboard in car to out the front, OK. But if you go from dash to rear view to out the front…takes time. Same thing happens with convergence in 3D – takes time if way different
-Registration – moving left and right eye views horizontally to move the whole scene closer or further
-Depth grading – adjusting explicity depth fo objects by shaping and moving individual objects or groups of objects towards or away from the viewer
-perfect 3D – can’t type that fast
-this is dave from InThree
-if you’ve got shots that are “at” the screen, can adjust to move it behind the screen – but your stuff that was almost at infinity in background is now beyond infinity, but causes eyes to diverge. So make darker or defocus to get it done
-25% depth budget on a TV. If you re-register, if you up that range, you can re-pdeth grade
-they tested stuff for 40 foot screen, 15 foot screen, 5 foot scren, showed it to folks. perferences when viewing ona 5 foot screen content that has been depth graded for:
-5 foot screen – 14% had experienced eyes, 17% had inexperienced eyes
15 foot – 66 and 42
40 foot screen 20 and 41
Samsung HDTVs are pre-built to do 3D.
-gorgeous 3D on television should be graded for 15 foot screen seems to be the take-away.
Linnie Elbert – subtitles in 3D. “What’s the deal? Once you get into it, makes you throw up in no time.”
Polar Express had subtitles, had’em projected on a separate projector on the screen.
most 3D content has been children’s movies, are usually dubbed
-Beowolf was to be subtitled, was only dubbed in 4 languages
-started some tests, picked two methods to put the movie out – in 4 territories (Neitherlands, Japan, Luxembourg) – subtitles in 3D pulled out in front, action behind usually. Can’t do it in projector, have to create them. Have to create a brand new master with subtitles, often 15 to 20 languages, hence 15-20 masters. Ugh, untenable
Also tried to put’em above the image – Beowolf was 2.35, so put in the 1.85 box and shoved it down. They tried it both ways. There’s more action at bottom than top. A floor, or water, etc. More often, the top is at a vanishing point and is far away, less parallax between what you’re reading and what you’re trying to look at.
if trying to read it, there’s other content your eyes are trying to understand, is a conflict.
-PUTTING THEM WITH A LOT OF HEAVY BLACK OUTLINE/SHADOW SEEMS TO LOOK BEST
-John ??? worked on the post/editorial stuff on 3D.
-identify challenges for 3D post
-with an 11 week post for a 3D movie, that was a PROBLEM. “My advice to you is that if someone comes to youwith an 11 week schedule, run the other way.”
-the support of production editorial
-in a 2D movie, you can evaluate the set cut to director’s cut
-but for 3D, is imperative to keep close, conform in 3D for the editor as he was cutting a song a day, had to show daily to have an accurate presentation of whether his cuts are working or not. On a 7 camera shoot, he’s going through and editing one eye.
-when conforming, just matching the creative cut to show in 3D was NOT ENOUGH.
-cameras were on cranes, moving, etc.
-the convergence points were all over the place
-without convergence matching, was painful to watch
-can’t tell if the cut was OK, unless can see 3D, because the cut migth be OK, but the convergence might be off (a new variable to work with in editorial)
-digitize left/right eye, conform, do a basic convergence balance pass
-some beam splittter issues came up time to time, to make l/r comfy to watch
-convergence issues for first pass, a balancing pass, the most important new thing
(Hanna Montana) – Michael woudl cut, quick conform & convergence pass, then show it, make changes, etc.
-by time got to a “latched” cut (Is there a locked cut anymore?), a new set of challenges.
-the real work begins, a combination of doing the color correction, doing the real 3D convergence pass (not the quickie), and doing then in tandem – how they interacted with the 3D, gotta be playing with both to get it to work right
-as editorial is continuing to make changes, those choices affect what the convergence/colorist stuff is working on
-bought a new Quantel Pablo, new storage, all new gear, using development code, not even beta, all this brand new pipeline to do it, a new challenge
-in the future, stronger support for being able to do 3D evaluations, but is a while off before it happens
-the concept of working with conforming as you go, is something they’ll have to be doing. Is not a good idea to try to make 3D evaluations in teh cutting room, the director/editor want to see the 3D really quickly, to see how the 3D is affected by cuts, and how it is working
-is diffferent for every shoot
–4 1/2 minute music video – can tailor the convergence, have it pop, work it/push it more. For a feature, with 90 minutes, can’t work their eyes as hard. Can hit’em over the head for 4 minutes, but can’t do it for 90 minutes. Tailor your convergence for what is appropriate for the material and the duration
-graphics and titles – was it easy or hard? There is a learnign curve for the vfx/graphics houses, did some back and forth, supported as they could with title placememt etc., but on Missy Elliot 4.5 minute video, they had some floating graphics to get the hang of where to put the graphic in z space?
Do we have people up to speed on 3D graphics? No, is a learning process
-they built a 3D slate, as they started doing titles, “do it like the slate” that was doine in 3D – ironic that was how the standard become,
Q: -If workign off a 20 foot screen, safe for most of the widest releases, or do need to compensate for each of the deliverables. A: if designed for 20 foot screen, will work on 5, but parallax doubles on 40 foot screen, triples on 60 foot, so that’s bad. But you can’t tell till you can see it. So they may need a 40 foot screen to test the footage. GOTTA EVALUATE with a 40 foot or bigger screen.
Atttempted to do rendered in grpahics for some things – Benelux and Japan were the two territories – this is NOT a way to burn-in stuff, isn’t feasible to do it with the turnaround available.
On Hanna Montana Q: do you work on the ocular separation on camera, or it is fixed, how much is done set up in front, or how much later?
A: interocular set on set as shot, the controller they have is convergence control. As part of storytelling, you’re using/tailoring 3D. One thign to NOT have happen is to have convergence at 40 then 15 then 60 feet – is way hard on airs.
Ramp up to 3D moments, is a convergence balance, not interocular. Vince Pace did static vs. dynamic convergence when shot. Somebody controlling convergence on SOME camera rigs, but not all cameras on all shows
Q: -how do you monitor convergence on set?
A: had basic dual projection feed, could do it on set, most guys in consoles, the guys that drive it, could do an A+B or a differential between left and right, will show the convergence error – that is how the operators did it
Q: LaserLight engines – brightnesss – based on experience so far, how bright WOULD you want it?
A: no answer yet
Want to see 14 foot lamberts if they could. Hanna had 4.5 foot lamberts. Creatives wanted to push it further, went to limits of projectors. Mo is bettah.
Q: Waht about dual vs. single projectors?
A: dual projectors would take another hour – great for making more light, but HORRIBLE to keep them aligned. Keeping the geometry, focus (across time), and something else aligned, it’d be great, but those are all hard. Dual screen is A solution, but is it a long term solution? Dunno. Everyone is doing 2 color passes – one for 2D version, one for 3D version, and the 3D is dimmer, human eye can’t adapt, gotta do it both ways. Is something is at 5, 10, or 14 foot lamberts, is REALLY expensive to grade one for each type fo projector
Q: Convergence changes with age- it changes over time from young to older – have found difference audience reaction based on age?
A: Hard to say, as some older folks sya they’ve experienced about some issues young folks don’t say. Probably an anecdotal thing. People NEW to 3D have a hard time locking in. Folks who’ve seen it repeatedly can lock pretty quickly even on some pretty bad systems. Speaker interested on what he calls “Settling time” based on that dashboard experiement – if making movie for older audience, will motion through Z space have to be more carefully, gradually, softly dealt with?
Q: stereoscopic 3D – does it change the pace of storytelling?
A: Overdoing 3D – can’t do stacatto cutting? NOt really true. How to control convergence and set it appropriately for the action. For fight sequences, can cut fast, but CANNOT have convergence flying all over, because of the time to settle your eyes issue. Howard expected couldn’t have any slam pans – but you can, it works, can break the “RULES” so long as pay attention to the convergence on the Z Axis
When re-dimensionalized establshing shots, wanna look at different parts in scene – “savoring” the scene. 3D is unique and gorgeous for that. Peter O’Toole crossing the desert in 3D would be gorgeous. in the Matrix- slow bullets coming at you is gorgeous – can emphasize those storytelling elements in 3D
Howard – we have found that the 3D cut is different for the 2D cut. With 3D, wanna linger on stuff, as the brain has a lot coming in – if in 2D, can be relatively boring in comparsion – don’t want to savor/linger. Can’t just pull out one eye of a 3D movei to make a 2D movie -might not be optimal? Still so much more to be done to figure it all out.
Sessions start at 8:30am, reception across the hall…
-mike out to DRINK!