I am very excited about a larger project which may be coming together. I'm looking for suggestions regarding a 3D pipeline. Essentially, camera package + data (i.e. 1Beyond Wrangler, codec) + edit (red rocket, aja, card) and editing platform.

Love to hear what people are using.

Hey Jesse, good questions.

Since I'm trying to put together a feature-length doc, myself, I'm also very interested in what's worked for others. I'm still largely in the R+D phase myself (particularly in regards to animation), so I haven't made my final decisions yet, but I do have a loose plan.

When I made my first stereoscopic short, last winter, as a warm up for the feature, I cut it on Vegas. But I can't really imagine doing that with the feature, because Vegas is so unstable, unreliable, and clunky in procedure. Besides which, I've been a professional Final Cut editor for many years, so that system made the most sense for me. I have not yet invested in Dashwood's suite, just because my editing thus far has only been of the interviews, and I haven't yet needed to see any of it in stereo. I'm still using Vegas for cobbling together montages of stereo stuff I've shot or animated; it's a decent tool for simple edits.

When I shot the interviews back in April, I used my recently bought NX3D1 and a borrowed 3DA1, mostly because twin-lens cameras were what I understood at that point. Now, after playing with my side-by-side rig for a couple months, I would rather have shot the interviews that way. My live-action and CG stereo backgrounds are much deeper than the greenscreen interviews that will be composited into them, so the researchers will be somewhat cardboarded. Thankfully, since I'm creating a mind-trip documentary instead of a drama, I think I can get away with depth that is compelling, even if it is not quite naturalistic.

My current side-by-side rig is just two cheap unsunc HD cameras on a bar; I just make sure I clap in front of them, and use that as sync. That's been fine for my experiments in wide-IA shooting, but I'm hoping to invest in a couple of Sony FS-700s, for their high frame rates and fabled eventual 4K firmware update. Even though my interview footage was shot in 1080p, I'd like as much of the backgrounds and b-roll to be shot in 4K, as possible.

I've invested in a Glidecam, because I think 3D wants to be in motion, to show off the changing perspective. However, I've realized that wide-IA shooting increases the side-to-side wobble, because each camera is offset on a long lever, so I have to improve my steadicam skills, or settle for just using the NX3D1 on the rig.

As you've probably noticed, I'm learning animation on 3DS Max, using David Shelton's 3DHippie Stereocam rig. This has been great for creating impossible virtual sets, particle effects and text treatments that still have all the thereness of stereo.

And finally, I've been compositing everything on After Effects. I do some pre-comping in 3DS Max, by inserting animated bitmap versions of my interviewees into the virtual environment, and then rendering out the different depth layers in various render passes, so that I can align everything in AE, with the minimal ghosting possible.

I'm looking forward to hearing what others' long-form workflows are like. :)

I'm putting a long stereo 3D documentary together at the moment.

It was shot using a 3DA1 (on tripod, a shoulder rig and a jib) and a Z10000 (on tripod, shoulder rig and a Glidecam 4000). For most of the shoot I've used the 3DA1 recording into two Hyperdeck Shuttles, which works a treat although the limited on board battery life is a bit of a pain on location outside. I'm looking at some solutions to this for future shoots.

I'm going to edit it all on Edius 6.5. I hardly ever use FCP now, in spite of Tim's excellent plug-in, because Edius is just so much faster and easier for stereo editing than FCP - unbelievably quick in fact. I had to go back and use FCP a couple of days ago on an old project and it took me about 2 hours to make some minor edits: I did something similar in Edius and it took about 15-20 minutes.

I don't have any animation or CGI capability myself (I keep meaning to learn but haven't had time) but I have friends who are very capable in this area so if I need any fancy 3D titles or anything I will go to them. They usually work in After Effects for this stuff, which one day I will buy and learn how to use.

I've shot green screen with two other companies where they've put CGI and effects into the film, one was a short 4D cinema sci-fi film and we plan to make some more of these.

I have a bunch of other stereo 3D work in progress, mostly short films and educational videos all less than about 10 minutes in length and I am using the same workflow on all of them. I use a couple of Panasonic HPX250's on a Hurricane rig, either in parallel or mirror, for the more complex shots although for the current documentary I never felt any need to use it. There are some shots I have planned for upcoming work that will need it though, mainly wide IA for long shots outside.

I'm hoping to try out Ken Burgess' wide angle adapters and base extenders soon, very excited at the possibilities of extending the useable range of the compact 3D cameras.

... but I'm hoping to invest in a couple of Sony FS-700s, for their high frame rates and fabled eventual 4K firmware update. Even though my interview footage was shot in 1080p, I'd like as much of the backgrounds and b-roll to be shot in 4K, as possible...



Matt, the FS700 does not have genlock. How do you propose to sync two of them for stereo?

Matt, the FS700 does not have genlock. How do you propose to sync two of them for stereo?
Good point, Neil, thanks for pointing that out. I first noticed the FS700 cuz Sony was positioning it as useful for 3D rigs, and so I always assumed. Hard to imagine a $9G camera, with no genlock.

Do you know anything about JVC's low-end 4K camcorder, the GY-HMQ10U? I'm looking at its specs, but it doesn't mention genlock, one way or the other. Fixed lens, but only $4500, without the need for firmware upgrades or external recorders (records to 4 @ SD cards). Doesn't have the FS700s brilliant hi-speed bursts, but it does shoot 4K at 60fps. (Actually, technically, it's slightly less than 4K, but it's still 4X1080p). Because it breaks up the sensor image into 4@1080p streams, it's able to capture ~144MBpS.

I've also been following Cesar Rubio's home-made MVC (machine vision camera) stereo rig, utilizing 2 @ Falcon2 Color 12M heads, which record 4K at up to 58fps. His recorder captures up to 270fps in 2K. He seems very happy with the rig, but I don't know how well field-tested it is. It might be a very big investment into a tool with limited uses.

For the short term, my focus is on creating a promo for the documentary, so I'm fine with 1080, for now. But if the promo is able to net me any investors, I'd like to step up to 4K for the feature, keep it theater-friendly. Hopefully, by the time I'm ready to invest, the choices in 4K will have expanded, and the prices come down. Maybe, by then, Sony will introduce a version of the 700 with genlock!

The NEX-FS700 has a LANC port, so maybe you could sync the two using it.

JVC's little 4K camcorder does NOT have genlock either, something I spoke to the JVC folks about at length at BVE in the UK.

I find this very frustrating - AF101, FS700, BMCC, HMQ10U, all great cameras and not one genlock port between them. Why not??

The JVC does have 4x HDMI outputs though, so you could hang 4x Hyperdeck Shuttle or 4x Ninjas off it and capture at 4:2:2. Just not in stereo.

I think Cesar has the right approach actually, instead of trying to squeeze two massive clunky cameras close together on a single rig, just put the heads on it and move all the recording and control electronics off-board. This should yield a much more flexible solution. To be honest, all I want is something that outputs the best quality video stream to a couple of Shuttles or Ninjas, you can remove all the other recording-related stuff from the camera itself. Won't do the frame rates that Cesars system will do though.

The NEX-FS700 has a LANC port, so maybe you could sync the two using it.
Interesting point. I wonder if the LANC would make any difference, when it came to the high-speed bursts? The slo-mo is such a strong selling point for the FS700, and it would look so good in 3D, that it'd be a shame not to be able to use it.

JVC's little 4K camcorder does NOT have genlock either, something I spoke to the JVC folks about at length at BVE in the UK.

I find this very frustrating - AF101, FS700, BMCC, HMQ10U, all great cameras and not one genlock port between them. Why not??

The JVC does have 4x HDMI outputs though, so you could hang 4x Hyperdeck Shuttle or 4x Ninjas off it and capture at 4:2:2. Just not in stereo.

I think Cesar has the right approach actually, instead of trying to squeeze two massive clunky cameras close together on a single rig, just put the heads on it and move all the recording and control electronics off-board. This should yield a much more flexible solution. To be honest, all I want is something that outputs the best quality video stream to a couple of Shuttles or Ninjas, you can remove all the other recording-related stuff from the camera itself. Won't do the frame rates that Cesars system will do though.
What a shame that these cameras don't have genlock.

Yeah, I agree that Cesar's approach sure sounds good. The heads on the Falcon2 are really tiny, so applications like 3d macro, steadicam, etc. should be much easier to pull off. I'm just wary of investing so much money in a system that may not have the kind of support and service that a larger camera company would have.

Interesting point. I wonder if the LANC would make any difference, when it came to the high-speed bursts? The slo-mo is such a strong selling point for the FS700, and it would look so good in 3D, that it'd be a shame not to be able to use it.

I don't know how you genlock non-standard frame rates on cameras - not saying it cannot be done, just don't know how it works.

I doubt LANC could help in this case at all. Generally the two cameras don't digest the LANC commands at exactly the same time anyway. The only way the LANC helps is by powering the two at the same time, which is a hardware function, and therefore nearly assuring that the vertical syncs start simultaneously. Of course there is time drift due to mismatch of crystal frequency. Once one would change the capture frame rate on one of the cameras, all would be gone.

Speaking of frame rates, did you see this? ‘Hobbit 3D film wizardry left us feeling sick’ | Mail Online 423 Digital, Inc. (http://423digital.wordpress.com/2012/12/02/hobbit-3d-film-wizardy-left-us-feeling-sick-mail-online/?goback=.nmp_*1_*1_*1_*1_*1_*1_*1_*1_*1.gde_51511_member_191872455)

Maybe has to do with triple flashing vs double flashing: http://www.nbcnews.com/technology/technolog/new-3-d-movie-technology-are-viewers-ready-1C7206887 .

What do you think?

When you use consumer camcorders based on AVCHD, the use of lanc controller like the ste-fra-lance will help you to sync the cameras. To be more correct, it helps you to start the cameras in sync. BUT there are still two issues:

- if the cameras have not been paired in a good way, they may shift out of sync very fast. There is no way to control syncronisation with lanc, as one can do it with genlock. But after pairing consumer cameras = selecting two cameras that show a similar sync behaviour, you can shoot up to 30 or 40 minutes before the run out of sync and need a restart of the cameras.

- the startpoint will be still different by 0 to maybe 3 or 4 frames, and the endpoint will be too. So the clips will have a different duration. One has to bring that in line in the timeline. That is something that can be done in a nice way in Vegas, since Vegas supports the editing in the timeline in a great way - and it supports the pairing ot L and R in the timeline too. It can also be done in Edius, BUT you cannot pair two tracks in the timeline, only in the bin structure of Edius 6.5.

Point 1 - I agree, that's what I said, however, if the camera doesn't provide genlock port, LANC is one way to get better skew than just free running the cameras, that doesn't require modifying them. I never do takes long enough that the cameras (quite randomly picked) wouldn't stay in sync for their duration. I never need 40 minutes. restart the cameras once a while is a pain though. For the more adventurous, one just needs to replace the crystal oscillators with better matched ones, but that requires reworking of the cameras. I still don't understand why he semipro cameras don't have genlocks on them. These are useful for other things than 3D, and aren't that difficult to implement.

Point 2 - I have seen as much as 12 frames of initial skew. It strictly depends on what the F/W is doing at the time on each camera, The AVC compression complicates it further. Just about any 3d editor that I tried allows aligning on the timeline. I have certainly aligned the clips in Edius 6.5 on the timeline, but I had to go back to the bin to pair them.

I have certainly aligned the clips in Edius 6.5 on the timeline, but I had to go back to the bin to pair them.

Unfortunately, that is one of the major disadvantages of the Edius 6.5 workflow for s3D. Edius has a great quality and a great stability - but to allow the pairing in the bin only is a pitty.

I don't see it to be a big deal. It is no different than what I had to do with the older version of stereo3Dtoolbox, where I had to trim the start, generate subclip and pair the subclip with the respective pair member. This is only necessary with non-stereoscopic cameras. 3D cameras are of course aligned, so again, no big deal, to me anyway. I have to so lot more to the material during the edit session before I am done. At least I don't have to wait forever for rendering to get done, or loose ton of work after crash - which happens to me a lot using Vegas. Furthermore, Edius is compatible with just about anything I throw at it. I see no "major' problem with the Edius or with the S3D toolbox. Vegas crashing is very time wasting, however.

I do not see a lot of crashes with Vegas Pro 12 anymore. From time to time there are crashes, but that seems to be gone - at least on my system.

I see following drawbacks with Edius 6.52:
- the pairing issue as discussed here. In Vegas I adjust and pair the clips in the timeline - so for a full project.There is no need to bring back all the clips in the bin, make sure that you have them in the right sequence and that L and R is in the right sequence. And there is no need to delete the original clips after pairing from the timeline, and bring back the paired clips to the timeline again. I start in the timeline in Vegas, adjust all the clips in the timeline, pair them and I am fine.
- the issue that you cannot use the s3D plugin on the total timeline in Edius. You have to use that on the level of the clip, or on the level of the sequence. In Vegas I make my editing, and then use the s3D plugin at the track header - so for the full timeline. I set keyframes and adjust the whole project. With Edius that is not possible.
- the third issue I see is that there seems to be a small bug in the Edius s3D plugin, in the combination with rectify and the horizontal adjustment. When I adjust in Edius the horizontal adjustment, after doing that there is the need to autotrimm the movie. But when I autotrimm the movie, the horizontal adjustment is gone or changed at least. So I adjust again the horizontal adjustment.... that is an issue that I have seen in Edius 6.5, I am not sure if it is still there in Edius 6.52. In Vegas you do not see that issue.
- I miss an MVC-encoder in Edius
- I miss the authoring of 3D-MVC-based Blu-rays in Edius.as we have it now for Vegas with the DVDA6.

Beside that, Edius has some significant advantages:
- it is rock solide. A crash of Edius does not happen really
- it has a superior quality, the codecs are the best what I know
- Edius has added the capability to create AVCHD 2.0/3D structures. Means, that one can generate 1080 60p/50p for 2D, but also 1080 50i/60i for s3D structures that can brought to Blu-rays
- beside the bug in the s3D plugin, Edius is similar to Vegas with the capabilities of the s3D plugin.
- Edius is able to edit JVC TD1 files in s3D.

Both tools are superior for s3D I think. What one take is also a question of taste. I have here both, and like Vegas as my first system, and Edius as my second system.

The pairing in the bin is not a particular drawback, it's just they way it works. I did it today on a project and it wasn't particularly onerous but I must admit 99.999% of what I shoot and edit is genlocked so it's not something I worry about much.

I do like the slip function in the Onsight OS3D stereo plug-in, which allows you to slide the left and right clips relative to each other in 1 frame intervals after you've paired them without messing about on the timeline. I've not seen this in any other editor and I wish Edius had it. If I had a lot of clips to align I'd probably go back to FCP and use OS3D just for this.

Likewise I have no issue with stereo adjustment at clip or sequence level and would naturally do that rather than across a whole timeline, which could be pretty long. Most stereo corrections are at clip level anyway, only fine tuning for transitions at top level perhaps but I can do that in Edius later in the project by making a single sequence.

Normal flow to correct and edit a clip in Edius for me is rectify & fit, then autotrim, then play with the HIT if necessary, I haven't seen any particular problems with doing this and it's become almost automatic now.

I do wish there was a built-in blu-ray capability but I can at least export blu-ray format files out to an external authoring/encoding system (my friends have Adobe). Let's not forget that blu-ray licensing costs from Sony are one reason software to do this is so expensive, something they perhaps don't have to worry about for their own in-house tools. However this alone is not enough to convince me to move to Vegas.

As you say though, it's mostly a matter of personal preference but it is good to hear what each system can do and be able to compare them.

Well, for me every additional workstep is a drawback - especially if it is not necessary really. There could also be a pairing in the timeline. So frankly spoken, a workflow that forces me to additional steps is something that I do not like. Due to additionl work and the wish to speed up in a routine process, due to the possibility for errors. If you use genlock, it is no issue for you - because then you have no need to adjust the clips and then the possibilities in the bin are enough. But imagine people who shoot with consumer formats like AVCHD - they have the double work and I can tell you that I am not happy with that.

If one likes to do s3D adjustment at the clip level or the timeline level can be seen as a question of taste. Both has advantages and disadvantages. To do it in the timeline allows you to focus better to jumps in the depth, for example. But if one needs to copy clips he will lose the connection to the s3D adjustment that was done in the timeline.

Well, that is an interesting question - after the horizontal adjustment there is a need to crop the picture, otherwise you have 2D parts left and right in the video. I think the horizontal adjustment is one of the major editing operations in s3D - at least if one shoots with parallel cameras. So if the autotrimm dis-adjust the horizontal adjustment, I see that as bug. But maybe that is gone in 6.52.

As far as I know, there is no MVC encoder in the Adobe products too. There was an MVC-encoder in Edius during the public beta phase, but they have killed that again. Maybe one reason were the licesing costs, the other one maybe the very poor quality that we have seen there. In terms of costs - the upgrade to a 0.5 version was expensive enough. I think more consumer tools like Magix Pro X4 have an MVC-encoder, from Mainconcept and also Cyberlinks PD10/11 has one. I wonder why Edius charges us something about 300 Euros for the upgrade of a 0.5 version without such an encoder. Sure, people who use the Scenarist or DoStudio will not need that - but that is something about additional US$ 20.000. But not everybody goes for MVC-based 3D-Blu rays.

I do not see a lot of crashes with Vegas Pro 12 anymore. From time to time there are crashes, but that seems to be gone - at least on my system.
That's good to hear. I cut sometimes on Vegas '11, and it crashes all the time (especially when trying to render).

Well, for me every additional workstep is a drawback.
That's a good point, Wolfgang. I had no idea, when I started working in stereo, a year ago, how much extra work was involved. So many parts of the process involve extra steps in transcoding, aligning, pairing, etc. I'm sure we all get kinda used to it, but anything that can help reduce the overall time drag is very welcome. I'm looking for useful scripts and macros for After Effects, to help ease the repetitive extra steps.

Speaking of which, if anyone else here works in 3D Studio Max, you should know that David Shelton, the author of the 3DHippie Stereocam plug-in, has finalized his 2.0 release, with lots of great new features, and minus some old buggy ones. He has a pro version (which I was happy to invest in), but even the free version has lots of great functionality.
3dHippie Blog Archiv 3DHIPPIE STEREOCAM V2 AVAILABLE (http://davidshelton.de/blog/?p=311)

That's good to hear. I cut sometimes on Vegas '11, and it crashes all the time (especially when trying to render).

Right, there has been significant issues with Vegas Pro 11 - the latest build 700/701 should have improved much of that. And also Vegas Pro 12 has become much better, after the semi-public beta-testing. I do not say that it is now as rocket stable as Edius is, but you can work again without crashing every second minute. I see here long editing sessions wher it does not crash anymore ... not for all users of all machines, but it has become much better again.

I had no idea, when I started working in stereo, a year ago, how much extra work was involved.

Right - I should correct my statement in the way that I should say: additional steps, that go beyond the workflow that is optimal for achieving my goals, should be avoided - since they are additional work, reduce our productivity, and bring in additional possibilities for errors. s3D requires additional steps in the workflow, for sure. But with a black-and-white movie we also had no need for color correction - what is not true, since also that must be adjusted for the total luminance.... :))

But I think you know what I mean here. I want to be able to run a workflow that allows me high performance deriving by a good standardisation. I do not say that Edius is worse here. Overall, I think the Edius workflow is great, and I love it because it is a very professional and very stable tool. But I was shocked a little bit when I started to understood the restrictions in pairing of AVCHD s3D files in Edius. But I undestand very well: if somebody works in Edius, he will be willing to accept such limitations. A 100%-perfect workflow is an illusion in the end - but I would like to go for something of 90%.

I have no experience with Davids tool, I would like to learn here more from your experience.

I want to be able to run a workflow that allows me high performance deriving by a good standardisation. I do not say that Edius is worse here. Overall, I think the Edius workflow is great, and I love it because it is a very professional and very stable tool. But I was shocked a little bit when I started to understood the restrictions in pairing of AVCHD s3D files in Edius. But I undestand very well: if somebody works in Edius, he will be willing to accept such limitations. A 100%-perfect workflow is an illusion in the end - but I would like to go for something of 90%.
Yeah, I get it. We are exploring a niche that is not well-supported, and which is growing constantly. That means lots of great new products and workflows will emerge, but it also means that the infrastructure is not quite there yet (at least not on the low-cost-end).

All of us who choose to explore s3D need an adventurous spirit to try out all the possible ways of approaching it, but at some point, it'd definitely be nicer not to have to do all the manual monkey work, for each and every project.

I have no experience with Davids tool, I would like to learn here more from your experience.
Yeah, it's been great, and it just got a lot greater. His site (link in my previous post) explains all the functions better than I can, but his 2.0 version is much better than the previous script. For example, it now has a dedicated stereo window. In the previous version, the stereo was always the last thing I'd deal with, because the script could only render a preview of one frame, and it would take a long time to do it. But now the plug-in supports a second 3D monitor, so I can visualize the scene in stereo, the whole time.

Plus, if my chosen IA doesn't look as good on the render as I'd like, I can adjust the depth, according to whichever eye I'd like to keep, so I only have to render the other eye out.

I can choose pixel parallax, so no more trying to measure the screen. It can render with horizontal overscan, so I can adjust convergence without losing my scale. It can integrate with Wimmer's Stereoscopic Player, to preview there. The rig has automatic near, screen, and far plane indicators, so I can set-up the cameras with reasonable accuracy before previewing anything. It integrates with the VRay render plug-in, for naturalistic lighting, etc.

There's probably more I'm not thinking about now. I've used his rig for every render, and so there's things I take for granted. If you have any specific questions about the rig, please let me know (and/or check out his site).

Yeah, I get it. We are exploring a niche that is not well-supported, and which is growing constantly. That means lots of great new products and workflows will emerge, but it also means that the infrastructure is not quite there yet (at least not on the low-cost-end).

All of us who choose to explore s3D need an adventurous spirit to try out all the possible ways of approaching it, but at some point, it'd definitely be nicer not to have to do all the manual monkey work, for each and every project.

Well, if you understand the editing of s3D with footage from TWO AVCHD or DSLR cameras as a niche, then you can say that this is not well-supported yet - by Edius. That includes every footage from two cameras, that cannot be genlocked during shooting.

If I have a genlock available as the 99% of Neil, you have not that issue. If you shoot with a MVC camera like the Z10K, you do not have that issue either.

At least I have to explore a little bit the link to Davids tool.