Great article and lots of interesting pictures. I look forward to seeing the video on GAC.
Can you elaborate on how you used Plural Eyes to sync the 30p footage with audio from the board? You mentioned that some clips stayed in sync, and others didn’t. But what happens in the timeline to keep the clips in sync? Does Plural Eyes slice up the clip and extend the subclip in and out points to maintain sync?
I’m very curious to see if the CMOS rolling shutter affected the drummer’s movements as much as I have seen it in other music videos shot with the 5D.
http://techthoughts.org/2009/09/19/wetnoodles/
Also, as someone who prefers mixing multicamera live to tape (as it were) I’m curious as to how you feel the amount of ingest (on three computers across a whole weekend), days spend syncing, cutting each song individually, laying it out, and making a final edited master- (at which point both would go to color grading, etc.) was “was considerably less than a traditional multicam show” and “The Canon 5D has allowed us to deliver better looking projects faster and cheaper”
Are you properly billing all the time & equipment usage for ingest and editing? Because an HDMI to SDI cable on each camera, SDI mixer, and a good live event director (with coms) would have enabled you to walk away with an edited version of the concert, in ProRes, as soon as you said “stop capture” on the computer. Take the money that post took, and apply it to the production, and you get that week of post back to work on other projects and make more money.
I’ll grant you that it doesn’t give you the week or more that was used to carefully edit shots from rehearsals and alternate show days, but a day or so of editing the multicam master with the rehearsal footage would have provided that.
I’m very wary that a week + of post is being touted as “money saved” on production when, in reality, it was still money spent on the total production- or money not properly billed.
Anthony Burokas
IEBA Communications, http://IEBA.com
Posted by IEBA on 03/04 at 02:08 PM
On the syncing with PluralEyes, all the clips were converted to 29.97 prior to syncing. The software doesn’t slice the clips or adjust speed to keep sync. It sort of seems to put the middle of the longer clips in sync so the start and end could drift. I have a feeling the PluralEyes creator might be working on something to compensate for this since the software is so darn smart. There all those 99.9% speed change solutions out there but since we were handing audio mixes back and forth from audio post making these audio speed changes scared me a bit. I know a Pro Tools rig can conform these types of frame rate speed issues but if felt better to me to just slips some shots here and there to make sync, especially considering the rehearsal that we added.
The question on the CMOS drum sticks is a good one. I don’t think it’s as drastic as that link but it might be there ..... this Twitter comment sums it up well:
@mjeppsen: From a convo with a DP about 5D rolling shutter “I guess people excuse the limitations because 90% of the time you want to lick the screen.”
Posted by Scott Simmons on 03/04 at 04:29 PM
Should say “some people excuse” because I abhor the distortions. It’s the reason I do not own a CMOS camera (save the one in my iPhone).
If my images are going to be the official record of something, I want them to be as true and accurate as possible, and knowing CMOS distorts…
Posted by IEBA on 03/04 at 04:40 PM
Anthony, as for your comments about live to tape ....
as an editor I WANT to cut the show from scratch. It’s well worth a week (or more depending on the complexity of the show) to recut from scratch. I don’t want to take away anything from live directors as that is an AMAZING art in itself but I feel that taking time to study the angles and dig into the footage can make for a more unique show. I don’t know the exact production numbers since I was involved in post but there is a lot of people and overhead to get that cut live to tape. You’d have to talk to multicam directors to see if they feel it’s worth money saved to go live to tape but I feel the edit on the back end is work the $. Plus you wouldn’t get this look in a live show that we were going for with the 5Ds since you can’t get them live to tape. Cleaning a live cut has never been as satisfying to me as a full edit.
As for billing, the deal we worked was acceptable. I can’t talk specific numbers but 21 hours of footage would have taken a lot of time to capture off tape. Probably less transcoding ultimately since there was not babysitting time as tape requires and a lot of the transcoding was the machines crunching. As and editor I want all 21 hours available when I edit so tape or digital I want it all. All angles, all cameras if at all possible since I want to look at them all. There are projects when time doesn’t permit that but when it does (and if I can work the time in such a way that it can) that’s what I like to do since I think it makes for a better end product. Since this was our first 5D multicam there was probably a bit more time taken on several aspects than might have been required as that often happens the first time you work with a particular format and workflow. That will only decrease.
Posted by Scott Simmons on 03/04 at 04:42 PM
Yes CMOS distorts IEBA, as do all recording formats.
Wet noodle drumsticks with CMOS, or nasty vertical light smear from a CCD sensor. There is no perfect answer because no single camera is perfect, the goal is to match the best tool to the job requirements/budget.
Excellent article Scott, I really enjoy seeing people embracing the advantages of new technology, and working around the limitations creatively.
Posted by Bighatcowboy on 03/04 at 08:46 PM
Scott, I wholeheartedly agree with you about the benefits, of editing the way you did on this project.
I just had issue with the _twice_ mentioned “money saved” as well as “faster and cheaper” which seem to almost be contradictory to taking all the extra time to study angles and dig into the footage to create a more unique show.
Faster, cheaper, better, Pick two.
You’re clearly saying it was better, and I agree, but then it wasn’t faster or it wasn’t cheaper.
And I mentioned the HDMI to SDI converters because that’s exactly how you’d be able to switch these cameras live. Throw in a 2-day rental of a new Telex wireless CellCom and you have the live event basically covered.
Posted by IEBA on 03/04 at 11:21 PM
Bighatcowboy, I agree that vertical tears on CCDs aren’t perfect, but they are more akin to artifacts, like lens flares, than distortion of the image within. My eyes also have artifacts when I stare into headlights or the sun. Just like my eyes see a flash a single, blinding moment of light, just like a CCD.
The article mentions and demonstrates the un-normal flash effects that CMOS chips deliver. Then there’s the rolling shutter distortion that affects anything moving (to a greater or lesser extent*). No matter if it’s a pan, or a drummer moving his hands up and down, motion within the frame is distorted. There isn’t really a CCD corollary to that.
Then add to that the lack of a low pass filter causing moire and aliasing artifacts (probably not much an issue with this subject matter) and the actual low testable resolution of the vDSLRs* and I have to say that I’ll take the vertical tear of CCDs (which was largely mitigated with the latest versions) over all of the issues in CMOS cameras.
I agree the look is great. Put a CCD in the vDSLR and I’m so there. But the key reason manufacturers went to CMOS chips is because they are less expensive to manufacture. They’re still trying to figure out ways to handle light gathering (back illumination is the latest trend) and ways to subsample the chips to provide video without moire (center crop / windowing on the chip in the latest cameras) and adding software after the fact to fix the flash issues (Panasonic HPX-300) and computer processing to fix the jellocam (techthoughts dot org/2009/08/12/cmosfix).
There is indeed a need to match the best tool for the job. For me, given all that, CMOS-based cameras haven’t been a best tool. I just wanted to clarify the 3-rd party comment Scott presented that “people excuse the limitations” because I, and others, don’t.
* links
dvxuser dot com/jason/CMOS-CCD
dvxuser dot com/articles/article.php/16
dvxuser dot com/articles/article.php/20
linking stripped because it wouldn’t let me post otherwise.
Posted by IEBA on 03/04 at 11:54 PM
Great write up, Scott. I always love reading about out of the box workflows.
Posted by .(JavaScript must be enabled to view this email address) on 03/05 at 01:32 AM
Anthony, we’re just going to have to agree to disagree on this one. As I mentioned I’m not the producer so I don’t know the exact cost of this workflow vs. a more “normal” way of covering a live show but these folks I were working with are live concert veterans so you can bet that if this type of show didn’t save money vs the more “normal” way then they wouldn’t have done it or be out shopping it as a viable way to cover a live show. Like any workflow it has it place and it’s certainly not for every event. I just can’t imagine that 7 camera guys walking into a venue with these under $5000 cameras isn’t significantly cheaper than a full live event multicam HD production even with some extra edit time. I’m simplifying a bit sure but we’ve done a near identical thing for years with DVX100s and HVX200s so you can bet that if it wasn’t cheaper than that wouldn’t have happened. That was using the cheaper and better. The EP’s comment is also referring to 5Ds in general and using the camera in this environment is just one other place where it’s making in-roads.
Also: “And I mentioned the HDMI to SDI converters because that’s exactly how you’d be able to switch these cameras live.” By doing this how do you plan to get the signal from the camera itself back to the switcher? Running a physical cable? That never would have worked.
Posted by Scott Simmons on 03/05 at 06:37 AM
The 5D MKII HDMI port offers 1080i out, which then drops to 480i when you hit record. Barring some HDMI breakthroughs with the Magic Lantern firmware, I can’t imagine why anyone would choose to use the 5D for a live switch.
The way they shot this concert is really the only way you could with a Canon HDSLR. I recently helped shoot a multi-cam 5D concert for Rascal Flatts in much the same fashion, for many of the same reasons (unique look, camera sensitivity, etc).
Kudos to Scott and Filmworkers Nashville for building a solid post workflow to deal with this unique challenge. Quickly turning around a creative edit on 20+hrs of footage is no small task, and I’m really looking forward to seeing the finished product!
-MJ
Posted by Matt Jeppsen on 03/05 at 07:47 AM
Great article, Scott! t reminded me of a very similar project I worked on back in 2002, shooting the rock band Garbage with what was the Canon 5D of it’s day…
...the Canon XL1, in Frame Movie Mode.
We used five XL1’s and two Canon Eluras, and did just what you talk about - every camera has a zone, try not to duplicate shots. I wrote an article about the experience in DV Magazine, but since DV’s archives seem to have evaporated, unless you subscribe to Highbeam it is a goner.
However, you can see a small clip of the show at this link:
http://tinyurl.com/27pku8
Apologies for the compression, but I wasn’t in charge of that…sadly.
I love this kind of production - I call it Multicamera Microproduction. I do several concerts a year this way, usually with choral groups, and the end product (although not being aired on GAC) never fails to please the audience. And no truck involved!
BAJ
Posted by .(JavaScript must be enabled to view this email address) on 03/05 at 01:23 PM
Great article, and very thorough. Just a few questions though. Have you used mpeg streamclip to transcode your files? I find the results jut about the same as compressor, but does it faster, even when I’m quick clustering, and I find that compressor crashes when I try to batch too many files at once. Also did you ever find at one point with the 12 minute recording limitation problem in that if all cameras started recording at the same time, wouldn’t there bit split second blackout spots between all cameras? Or did you start the cameras a few seconds apart from one another to make sure you always had some kind of coverage?
Posted by Thomas Wong on 03/06 at 08:20 AM
Thomas, I have used MPEG Streamclip a lot but not for these conversions. I have heard that it is fast though and i want to test it out. And I too have seen crashes on some installs of Compressor when dragging too many clips in. Droplets seem to help but it is frustrating. There was more babysitting on the transcode that I would have liked.
There was never a problems of a blackout with no cameras rolling. There were enough cams the operators were able to stagger their start/stops or just toggle in between songs.
Posted by Scott Simmons on 03/06 at 10:09 AM
Thanks Scott, great to know. and I highly recommend MPEG STREAMCLIP next time you work with 5d. It’s really quick to pro res, and you can batch as many files as you want without fear of crash. (less babysitting more time to go get coffee) My only problems with it are that I wish it would use more processor power like you can by quick clustering in Compressor, and or it’s inability to have multiple instances of it for more batching…
Posted by Thomas Wong on 03/06 at 10:30 AM
Just so you know, with Mpeg Streamclip, in the Batch List window, you can set it to work on 4 simultaneous tasks. I’ll have to set it up to test to see for sure if that uses multiple processors, but I believe that it does. It won’t split one job over several cores, but I’m fairly sure that it’ll split separate files withing the list up. It’s worth looking into.
Posted by .(JavaScript must be enabled to view this email address) on 03/09 at 01:20 PM
Yes, I do use that feature often. However, I don’t think it uses more processing power at all. I look at my CPU usage while doing so, and there is no difference to if I was only encoding one stream at a time. It spans multi-core, but only uses maybe 20% of CPU. It would be great if you could have less overhead with it and churn dailies/transcodes/masters etc. out that much faster.
Posted by Thomas Wong on 03/09 at 08:28 PM
I hope this isn’t an obvious question, but why transcode the h.264 to ApProRes vs editing from the original files. I have only played around with 5D camera tests in FCP but even in a software only setup I can playback the h.264 files. Would an alternative workflow be edit in the 30frame h.264 files and taranex the output to 29.97 for SD delivery?
Great article, can’t wait to dive in deeper on future projects.
Mark
Posted by .(JavaScript must be enabled to view this email address) on 03/11 at 03:01 PM
donmarconi ... several reasons to leave the H264 codec. It’s really a delivery codec and not a post-production codec. that’s what something like Pro Res is made for to transcode to for edit. While FCP will playback the H264 files it’s very processor intensive and once you start adding transitions, layers, effects, things grind to a halt. More importantly you can’t send the H264 files out a Kona card to a client monitor ... a necessity not just for the edit but to master to tape in the end.
Posted by Scott Simmons on 03/11 at 11:17 PM
Great article Scott, just what I’d been looking for. We’ve been shooting shows much the same way since 2003, starting on the XL1 and working our way up to the XLH1. The 5DmkII seems very interesting, if for no other reason (though there are many reasons), it’s such a better investment than any video camera. You’re putting your money in to glass that holds its value, rather than a video camera that will be practically worthless in 5 years. I have a couple quick questions though. Regarding the 4GB file limit, do you have to manually re-start the recording, or will it roll over in to a new file automatically? If it rolls over, how many frames do you lose? Also, what kind of shoulder mounts and/or follow-focus kits did you guys use for the shoot? There is quite a range of products out there. And finally, have you seen anyone use a 5DmkII on a large jib with full remote zoom/focus control? I’m looking forward to seeing your concert on TV tonight, wish it was in HD though, darn broadcasters. 
Posted by capturedliveproductions on 04/07 at 11:05 AM
i can chime in on a few answers if nobody minds. 5d will not roll over to next clip, it literally stops and you have to hit record again (don’t know if that changed in new firmware) so frames lost are as fast as the operator press record again. Main reason for so many cameras I believe, as Scott mentioned to my previous question, he never ran into any black holes in the edit.
Posted by Thomas Wong on 04/07 at 11:15 AM
