Fiddly to balance?
Just came across this puppy: the Hague DSLR Motion-Cam Stabilizer. It’s a very cheap handheld stabilizer and… well that’s it. It works with a polished ball/socket type of gimble. From the look of it the weight seems to sit over the hand which means it’s probably easier on the wrists than the Glidecam which will destroy your arm in no time at all. Having worked with these stabilizers a lot I know the key thing is how quick you can get balanced, and, to be honest, this just looks like a pile of gash. The Glidecam is a pain to adjust and this looks like it’s going to be even worse, but then you can purchase it for £127.62 ex vat, which is pretty damn interesting. Stabilizers are, I think, an absolutely essential piece of kit in your DSLR armoury so if you can’t afford the Glidecam then this might provide you with a chance to get to know how they work, for a tasty price. My instincts tell me that this will probably be rubbish, but I thought the Glidecam would be and it wasn’t. Cheap. Cheap. Cheap. Haven’t found any video shot on this little puppy yet but there are some pretty wretched efforts shot on other Hague products on youtube if you want to see them.
PS Hague really need to do something about their website, wowee that’s bad design.
There have been loads of comments on the 5DtoRGB post I did this week and it seems to have struck a chord with people looking to optimise that important stage of the process. The fact is, most of us are just guessing and don’t really understand what’s going on in the weird and wonderful world of Codecs and compression algorithms (myself included). Fortunately one of you, my beloved readers, actually does have a pretty good handle on this, so I’m posting his thoughts in this entry because it’s pretty interesting. I only know him as Mark but he’s a long-established editor and veteran of the bad old days of broadcast cutting. Enjoy, and thanks Mark for taking the time…
It may be possible that MPEG Streamclip (your comparison benchmark) has errors in how it decodes the H.264 data in which case a program that has less (or no) errors would give you “better” quality. Just to be clear though, in that case 5DtoRGB is not making the data *better* than what was in the original file, just *closer* to what was in the original file. There is a mathematically optimal way to decompress the pixels and then expand the 4:2:0 color data into a larger color space. The optimal method doesn’t retrieve (or create) data that wasn’t in the original file to start with. If it does its job perfectly, the best outcome will be that it doesn’t further degrade your image quality.
Any “improvements” you perceive beyond that would be the result of the software choosing to process the data differently or to add additional post-processing. Examples of this would include doing things like applying a different mathematical weighting to the interpolation of the data as it is mapped into the larger color space. However, to the extent that any processing is done differently or in addition to the optimal algorithms, the resulting video will be a less accurate reflection of what is in the original file. While this different or additional processing may have the effect of making certain kinds of footage more visually pleasing, it isn’t more *correct*. It also is very unlikely to be more visually pleasing on all possible kinds of footage.
Most importantly, it should understood that once the file is properly decoded and correctly mapped into a larger color space, any post-processing program (like Magic Bullet) or any competent three space color corrector should be able to achieve the same (or similar) visual results with no further loss in quality versus having the transcoder add the correction or effect during the transcoding process.
The disadvantage to doing any additional (ie different than mathematically optimal) processing in the transcoding step is that the processing is now baked into the resulting video shots. The advantage of not adding any “visual flavor” to the shots in the transcoding step and instead doing this enhancing in your editor or color corrector is that the changes will be non-destructive and reversible. This is why I believe that any decoding or transcoding software should stick to the optimal algorithms which are well-documented in a variety of technical books. 5DtoRGB may well be doing exactly that, which would mean that the improvement you are seeing is actually from eliminating errors introduced by your previous transcoder (MPEG Streamclip).
Keep in mind that if the format you are re-encoding to with 5DtoRGB is not mathematically lossless, then you are adding additional degradation and artifacts at that point. The absolute highest quality approach is to use an editor that works with your camera source files in their native format with no transcoding. This will always yield a superior result versus re-encoding no matter the method. Comparing other approaches is certainly interesting but ultimately it’s an exercise in selecting one inferior approach over another more inferior approach.
I’m always amazed at how willingly editors today accept a workflow that degrades every one of their shots by default. It is perhaps a necessary evil but we shouldn’t act like transcoding and recompressing to secondary intermediate formats is *normal* or even *acceptable*. It sucks and the industry needs to get past it as soon as possible. Someday soon all professional editing workflows will work natively with source camera files.
Anyone doubting that probably wasn’t editing back in the early 90s when the newfangled standard def “DV” format was bringing our state-of-the-art 486 computers to their knees. We had to transcode that DV footage into an intermediate format just to edit it non-linearly. Interestingly I remember everyone hating doing that to our precious footage with a passion that bordered on seething rage. Today people act like transcoding is fine. I think in a couple years we will be embarrassed that we collectively wasted so much time and quality simply to work around the “bug” of our editing tools being so primitive.
I think that H.264-based codecs will be around a for quite a few years yet. Compared to everything that came before it and also its current competitors, H.264 is really, really good. And I think it’s going to continue to improve for two reasons. First, it’s a difficult codec to implement correctly and the engineers are just now getting good at squeezing the most out of the base profiles. Second, there are several tricks in the advanced levels of the standard that the hardware based capture codecs haven’t even touched yet. I think we’re going to see quality per megabit get even better than it is.
I predict that what will replace pure H.264 will actually be a variation of H.264 because I suspect that more cameras are going to start capturing raw-ish video data (meaning that they might initially only retain part of the raw data due to size ). I also think that manufacturers will start compressing that raw data with lossy but good techniques based on H.264. While not as pristine as losslessly compressed full imager raw data would be, it will still be a significant leap over what we have today. The difference between the “semi-raw” I’m describing and “pure raw” will be one way that manufacturers segment the market to preserve their ability to capture big bucks from those that have it. The $50k and up cameras will have pure raw along with higher res (4k), faster framerates (over 60 fps) and higher bitrates (100 Mb and up). 4:2:2 will become more common in the lower end and 4:4:4 will be for the high-end.
With the move to raw video there is certainly a risk that the industry will fragment into a zillion custom raw formats like still photography did. There is also hope that either our tools will find a way to deal with all those formats (as Aperture, Lightroom and Photoshop do today) or that a meta-raw container format (for example Cinema DNG) will gain momentum and be widely adopted by manufacturers.
If I’m right, NAB 2012 should be a pretty exciting.
Canon’s been going completely bonkers, as is their wont, over the last week or so releasing details on all sorts of new stuff, and some which isn’t actually stuff, more just bonkers, like the ‘super super super duper wow oh my god what how did they what about rolling shutter then eh, er, what’s the point’ enormous CMOS sensor. Buried in all the announcements were a pair of teeny tiny, but very powerful, little camcorders, the XF100 and XF105. I tested the XF305 quite a few months back, before it was released, and it really is a very good camera, but the world has gone DSLR crazy and it doesn’t seem to hold quite the same allure as its flashier large-sensor whizzkids. So, why am I blogging about the XF105. Well, I’m embarking on my first 3D project, and need a really small, portable, smart system to shoot on that will give me the ability to shoot some properly inventive work. I read this interesting piece from Matt Zoller Seitz at Salon.com about what you might be able to do with 3D given a little creativity and I totally agree. Up till now I’ve completely discounted it, seeing it as a headache in a can, a fad, a cynical way to bump ticket prices, or create ‘added value’. I loved Up in 3D and there was no doubt Avatar was a deeply involving experience in 3D too. But I was recently given an opportunity to try something out so I thought, why not? The XF105 has a built-in assist function designed to help line up two models to work with each other. Whether we like it or not, our TVs are going to be 3D ready before we know it, and, in the same way that HD just happened, 3D will just happen too. Having humped my beastmaster (more on that in a later post, but it’s starting to get very heavy… ooh call me Vincent!). Redrock rig up and down a building site all day I quite like the idea of something neat, powerful and portable. So here, to finish are some pictures of me carrying the beast on a building site, replete with regulation YMCA helmet, raving gloves and subtle gilet accessory. Rock. And. Roll.
UPDATE: For whatever reason this seems to have touched a mahoosive point of interest across the web so I’ll be doing a much more thorough comparison on this issue once I get back from the shoot I’m on. Should be an interesting one, who knows what we’ll find!
I’ve been keeping an eye on this one for a while, Rarevision’s free 5D to RGB Beta which claims to get you
…as close as possible to the original data off the camera’s sensor while putting the brakes on any additional quality loss. In short it’ll make your footage look just plain amazing!”
Not to blow my own trumpet but recently most of my stuff has been looking pretty amazing on its own so how much help do I actually need? Well, maybe more than I think. Now, we all know about gamma shifts (why does it happen when RT is turned off, and not when it’s turned on in Final Cut….?) and H264 codecs being crap, and transcoding being a messy and time-consuming beast (yes Adobe Premiere users, I know you can edit native, thank you). I won’t go into the whys and wherefores of color-subsampling and why you need to be working in RGB not YCbCr because you simply don’t care. And nor do I, it’s really not that important. Just know that you really do. The programme works like any other encoding solution, giving you a simple interface to change your files into something you can work easily with, but apparently this conversion process is the best currently available. Big claim. They also say this:
5DtoRGB takes a no-compromise approach to quality, ignoring any concerns about speed.
Hang on… ignoring any concerns about speed? Ah… okay. I shoot on RED a fair a bit and believe me, the process of getting your rushes to the edit is so protracted and slow it’s agonising. Well, lots of people have been talking about this programme but no-one appears to have done any tests on it yet and I wanted to know: just how slow is it, and is that speed compromise worth it in the end? Well, I ran a test on a piece of footage from Ladies and Gentlemen, not a particularly challenging shot, but nice and wide, with plenty of people in the frame, lots of light variations, and it was shot at night so I’d expect there to be those tell-tale artifacts you get when the ISO is tickled a bit. The shot is 32 seconds long, or 810 frames. I transcoded it using the following settings comparing MPEG Streamclip, the current speed king, with 5DtoRGB.
Results? MPEG Streamclip finished the job in 31 seconds. 5DtoRGB did it in 4 minutes and 13 seconds. For those of you who like percentages that means that MPEG Streamclip did it in 12% of the time of it’s illustrious counterpoint, 5DtoRGB is 8 times slower. Hmm. Do the maths. MPEG Streamclip is pretty much real-time. You shoot three hours of rushes, three hours of transcoding more or less. 5DtoRGB you’re going to be looking at 24 hours. I shoot three hours of rushes on RED and I’ll be transcoding for three days, so it’s not quite there yet but it’s not great. I’ve seen other estimates that said it was around 5 times slower but I can only go off my own machine (MacPro Quad Core 2 x 2.8 GHz, 10GB RAM) which is, after all, what I work on.
So what are we getting for our machine time? Well, firstly, a few more options, the all important Chroma smoothing which, my instinct tells me, is going to be worth it. We also get an option to set the output gamma to a contrastier or flatter setting, my grading nostrils are now twitching.
As you can see from the settings I’ve used ProRes 4444 and the flatter setting to transcode the footage so we should be seeing the very best results possible out of the software.
And?
It’s an epiphany… Holy crap, this thing is unbelievable! The first thing you notice is just how insanely smooth the footage is, it looks alive, preserved and ready to grade. In other words it is in a completely different league to the footage from MPEG streamclip. First, it’s lifted the gamma and given me a flatter image and more dynamic range. Compare the grabs below. It’s also had the unexpected benefit of fixing some (not all) of the moiré issues, in our actress’s shirt for example. You can really see it in those close details around the edge of her shirt which are so jagged in the MPEG Streamclip version, but really protected (as always, not perfect, but dramatically improved) in the 5DtoRGB. Click on the images for the full size versions.
Wide shot using MPEG Streamclip, notice the moiré
Same shot using 5DtoRGB
MPEG streamclip scopes, clipped
5DtoRGB scopes, smooth as a baby's
In the test shots the company themselves posted they were zooming in 800% on the image to illustrate the effect and, in my cynical way I thought that it would just be a case of geek’s goggles (the kind of thing the naked eye doesn’t see, but that all the geeks will tell you is absolutely vital) but not at all. I test all my footage on terrible cheap 26″ monitors because there’s nowhere to hide. The MPEG Streamclip doesn’t hold up, but the 5DtoRGB is stunning. I can’t really explain how fantastic the transcode is. But believe me, it’s phenomenal. I worry deeply about grading my footage and I have two major projects about to go to grade – now I’m satisfied that they will be absolutely fine, as long as I put aside a couple of days for the hamsters to chug through the maths.
Detail from MPEG streamclip, compare the shirt and the light
Detail from 5DtoRGB, compare the shirt and the light
I’m not sure how the Vimeo file will render the difference between the two shots but definitely worth having a look. We’ve been needing this software for so long, and yes it does take a huge amount of time to transcode, but if you’re aiming at drama or television, or perhaps a cinema release then this goes a long way to easing some concerns in my mind. Well done Rarevision, they’re post specialists and it shows. No, it’s not going to sort everything out for you but my word is it good. It also gives you the option of exporting DPX files for effects work or for final grading in a Davinci suite or similar. I love that. MPEG streamclip for speed, 5DtoRGB for quality, and right now, they’re both free. That’s outstanding.
5DtoRGB test comparison with MPEG Streamclip from Robin Schmidt on Vimeo.
There’s a stack of new lenses coming from Canon, which is nice. Rob Galbraith has done a proper overview of them all, but the one that interests me is the 8-15 fisheye. That’s absurd. On a 5D it’s going to be wider than Pavarotti and only really suitable for daft hip hop videos and skateboarding, or is it? Anyway, lots of stuff coming out and I know there’s more in the pipeline. Photokina ought to bring lots of new stuff into the public eye at last. Also, the Canon Expo (an event that only happens every five years) is just round the corner, on the 2nd and 3rd of September. Sure to be some big announcements there. How do we keep up?
Do I want it? Yes I do.
Canon 60D - never hurt anyone, good for students
New camera, Canon 60D is kicking up all sorts of fights from people who declare themselves pissed off about it, that it’s a downgrade from the 50D, that it’s rendered the 7D obsolete, that it’s pointless, that it’s not worth $300 more than the T2i/550D, that it should have had a DIGIC5 sensor, that it should have had a flame-thrower and a webcam link to Phil Bloom’s underwear drawer. If it’s a downgrade from your 50D then don’t buy it. If you think it’s rendered the 7D obsolete then sell it, if it’s not worth $300 more than the T2i then don’t buy it, and if you think it should have a DIGIC5 sensor then wait till that camera comes out and buy that one, and if you think it should have had a webcam link to the Bloom’s knicker repository then clearly this DSLR craze has gone too far.
It’s just a camera, it sits in a range of cameras, and if you buy it then how you use it is up to you. You’re a consumer, you have a choice, and in this case the choice is an articulating screen and manual audio and 24p and stuff and well, just consider yourself blessed you have so many options, and if you think you bought the wrong one, then just remember the golden rule: invest in the lenses, buy the camera. You can always sell the camera on. The secondhand market for DSLRs has gone through the roof, absolutely gone through the roof. Check check. The 60D, if you read the press release, is aimed at students, it’s aimed at film schools, creating a product that will appeal to that market and give schools an affordable product. Yes, we shouldn’t read everything in a press release, but it just bugs me that so many people have got all upset about it when there are so many options available, not just from Canon, but from everyone now. Get over it.
So, for god’s sake, stop moaning about what the camera does or doesn’t have, we are so spoiled it’s insane and the next generation will spoil us even more. Take a valium, shoot some stills, shoot some video, or don’t and wait for something better to come along. It’s just not that big a deal.
So, here’s a fun game. It’s four o clock, you’re expecting a 70-200 USM IS 2.8L (yummy) for a job with a demanding client. Email comes and, for some reason, this lens is suddenly going to be arriving the next day. In other words, too late. Now, you could have put other plans in place if you’d known, but now you’re stuffed. Or are you?
So, I’m in Clapham and I need a 70-200 for the next day’s shoot. What do I do? Here are my options. York Cameras do great secondhand lenses, and since they’re near a bunch of other camera shops, that’s my first port of call. Half an hour to get to Tottenham Court Road, and it’s really raining hard, buy York Cameras are shut. Uh oh.
So, I’m faced with these choices:
The original lens I wanted 70-200 USM IS f2.8L for £2k
The 70-200 USM IS f4L for £1k
the Tamron 70-200 f2.8 for about £500
or a secondhand Sigma 70-200 f2.8 which i can get for £950 as well.
I have exactly one hour to buy a lens before the shops shut, and I absolutely have to have a 70-200 zoom lens for my shoot the next day. Which one do I get?
According to the guy in the shop the Tamron is soft. Ah. The f2.8 is £2k. Ouch. The Sigma is intriguing at that price but it’s not stabilised. The Canon f4 is known for being extremely sharp, sharper than the 2.8, and it’s stabilised, but it’s still a grand and it’s f4 which worries me as I’m shooting a lot of interiors without lights… I need stabilisation, I think at least, because I’m shooting video and long lens work can use all the help it can get on the DSLRs. I also want to spend as little as possible as this is coming straight out of my pocket when I really wasn’t expecting it. I’m going to have to sell the lens after the project as well so the less I lose on it the better. I was miffed to be sure.
So, which one did I get?
First correct answer wins nothing. Sorry, got nothing to give away apart from a witty putdown, which might not even be that witty.
AND NOW THE ANSWER.
The reason for even posting this is that I think it poses a very interesting question for a very interesting lens category. 70-200 is really useful, and I do find myself needing it a lot for video. I’ll whack the 24-70 on the 7D and it’ll give me a longer throw of course but I mostly shoot on the 5D now so that’s no help. My own wretched trial and error gauntlet with these cameras has led me to believe that there’s an exponential relationship between the sharpness of a lens and the deterioration in image quality because of that H264 codec. Factor in simple economics, I don’t have 2 grand kicking around to splurge on a lens, and it should be pretty clear which one I bought.
Yes, the Canon f4. We’re shooting a lot of exteriors, and though the interiors are going to be a stretch I couldn’t bring myself to buy lenses that a) weren’t as sharp, or b) weren’t stabilised. The IS on the f4 is a noisy brute, but you can really tell it’s doing its job. I have to sell this lens when then job is done, sadly, but, you know what the worst thing is? We didn’t even use it yesterday.
I’ve been trawling the weblet for videos shot on this camera and on the whole feel pretty underwhelmed. Some lucky chaps have got their hands on them and you can read a review here at Luminous Landscape, plus some pretty wretched video, though it doesn’t seem to suffer too badly from rolling shutter. I also found the clip below comparing the 5DMk2 and, the VG10 and the GH1. Know what I learned? Nothing. It feels like a knee jerk camera, and I’ve no interest in those kinds of products. History won’t care either!
It’s edging closer, the Panasonic AG-AF100 has its own website now and this time it doesn’t actually look like a 3D render in the photo. It’s not a particularly well designed website but a website nonetheless. Good on them for going for it, the first iteration of anything is always going to be troubled but we’re hopefully going to see a lot more innovation over the next 18 months from the other players. Canon notoriously play their cards very very close to their chest indeed. Who knows what direction they’re heading in, but it couldn’t hurt to do a large chip camcorder, could it?
Impero HDSLR Controller
Just came across this puppy. Viewfactor make lens control units for follow focus and zoom etc. and they’ve made the inevitable move into DSLR tech it would appear. The HDSLR controller does something I’ve been wondering about, making use of that complex USB port on the cameras which I’ve always thought ought to be able to deal with all sorts of interesting applications. And here we have one. Functionality wise you can control focus of Canon lenses through the USB port, no joke. It gives you a 10x zoom pbutton to instantly check focus in detail, a start-stop record button, and a focus wheel. Very nice, and of course no-one’s seen one actually working but we can at least assume that it’ll work reasonably well, given the company’s track record with their other products. Apparently it’s coming soon, and will cost around $499 for a wired version and $949 for a wireless version. That’s actually pretty expensive when I think about it, but we’re dealing with an ultra nice product here so you have to expect that.
Viewfactor also make a neat-looking powered cage which I would definitely be interested in for longer projects, like a feature film, where you need a solid device allowing for fewer cables and loose connections. Will let you know when I know more.
CONTINEO 5D/7D Powered Cage
