I was asked to do a presentation to some non-video types about the new developments in DSLR video last week. It occurred to me that explaining why the footage from these cameras is so nice is kinda hard unless you understand the concepts of Depth Of Field and the way the eye/brain views/decodes things. To make it easy on the audience I quickly shot and edited something with lots of shortDOF work in it and a little grading to match the footage and give it a suitable ‘summery’ feel.

Here’s the original one:

After the presentation (which went pretty well, they got the idea) I came around to doing a few exports of various sizes for my backups and deleting the raw footage to save space on the old laptop. As it was a nice short bit of footage I thought I’d have a go at sorting the annoying camera-shake that you get with shooting handheld which was only made worse by the flowers waving in the breeze.  After a little search around in FCP I found the ‘SmoothCam’ plugin. This is a basic version of what Motion offers in terms of video movement analysis and restoration. All it really does is try to identify constant still artefacts in the frame (a bit difficult in my flowers clips I know), then tracks them throughout the clip using the movement as a cue for stabilisation by moving the whole frame to counteract the movement of the camera. Obviously this will leave bits of the frame unfilled if the picture is being moved about to it will zoom in a bit to re-fill the blank bit of the frame (a bit of a downside, but if you’re shooting in 1080 and exporting in 720 its not the end of the world). You have to tell it to analyse the clips in your footage bin which takes some time, but once you’ve done that bit they’re always analysed and rendered so no matter what you do to them they’ll always be happy to play. The way I did it was to edit the piece then work out which clips I used then only analysed them as it takes some time. Probably the most streamlined way of doing this is to do your edit then make new masterclips of the footage you used and then analyse only them as it’ll be literally only the frames you’ve got in your timeline. I haven’t tried that way but it’ll be worth investigating if you’ve got time. But enough of the overview, lets see how its done step-by-step.

Step 1. Get the SmoothCam columns in your bin. Right click on the info columns in your browser and select ‘SmoothCam’ from the list.

Step2. Select the clips you want to analyse, right click in the ‘SmoothCam’ column and ‘run analysis’. Go get a coffee.

Step3. bring up ‘SmoothCam’ from the Effects tab. Its in Video Filters/Video.

Step4. Drag the effect onto each clip in the timeline that’s been analysed.

Step5. Select the clip that you want to work on as you normally would with dealing with effects and you’ll see the ‘SmoothCam’ options in the Viewer’s Filters tab. Have a little play, I finded it useful to deselect the SmoothCam effect, watch the clip, then reselect it and see what’s its done in default. Then you can change the ‘Translation Smooth’ (horizontal and vertical smoothing) and ‘Rotation Smooth’ (you can guess what that does) to see how much your clip really needs to be fixed. I found the default of 2.5 to be a little strong as I still wanted some of that hand-held feel to it. ‘Mix’ gives you a mix of effected and unaffected footage, ‘Scale Smooth’ lets you control the push-in and pull-out of the frame that covers up the blank parts created by the stabilisation, ‘Auto Scale’ will do that but automatically.

so what does it look like? Here’s thing run through SmoothCam:

And here’s the original with the 2.35:1 ratio for comparison:

Picture Style Settings For SuperFlat

Sharpness: 0

Contrast: -4

Saturation: -2

Colour tone: 0

Highlight Tone Priority: ON

ISO

160, 320, 640, 1250

FPS – Shutter Speed

24/25 – 1/50

50 – 1/100

50 to be slowed by half to 25 – 1/200

REMEMBER TO WHITE BALANCE!!

I shot the same exciting footage of pouring a glass of water on a 7D set to 720@50fps using 1/100, 1/200 and 1/400. I then converted it to ProRez422HQ and conformed the 1/200 and 1/400 from 50fps to 25fps in CinemaTools. I then stuck it all into FCP7 into two timelines, one 50fps and one 25fps. As there wasn’t a conform to 12.5fs option in CinemaTools to quarter the 1/400 I took the halved 1/400 footage and slowed the speed by half using FCP’s speed tool, leaving blend/scale/ripple all on.  Below are the two results exported as 720 H.264 25fps and 50fps.

50fps timeline

25fps timeline

I having looked at the clips a number of times, zoomed in on the detail, gone through them frame by frame I prefer the 1/200 slowed by CinemaTools to 25fps in the 25fps timeline. If you were going for a really slow shot I’d defiantly suggest the 1/400/CinemaTools/FCPspeed option as although a little soft, its still very much usable.

But that’s my opinion, and as with all these things its down to personal taste. So which do you prefer? Comments are just below. And if you’d like me to re-do the test in anyway just shout and I’ll see what I can do.

If you’re the owner of a 7D or 550D you have experienced some issues with Noise on your recordings thanks to the rather over zealous work of the AGC -Auto Gain Control, if you’re not familiar with this concept, its something that is designed to assist the camera in picking out sound when the volume going into the device decreases. On the 5D mark II and 1D there are hacked firmwares by Magic Lantern which can allow you to disable this feature of those models, but as of the time of writing, no such firmware hack is available to 7D / 550D owners.

This is where a hack comes in! Thanks to Mike B, aka Fony on youtube, you can effect a solution with the use of a simply Y connector (stereo to 2 mono) and a cheap mp3 player. Mike’s technique involves creating a false noise floor, which forces the AGC to hold back from turning up the mic, well enough yap from me, check out his solution in the video below -

http://www.youtube.com/watch?v=91xw_NX0Qm8

Some folks in the youtube comments have stated that they have had issues with this solution for the 550D, haven’t had chance to try it just yet but look forward to finding out if anyone has luck with it!

Came across these excellent videos from Adam Stevenson on Vimeo.com. I represent them here with his permission, cheers Adam!

Here you’ll learn some useful other potential picture styles, and in the comments you can find links to picture styles you can upload to your camera using EOS Utility (Canon only, sorry!). Just plug your camera in, set it to manual camera mode and then click ‘Register User Defined Style’ then upload the files Adam offers to your User Defined 1, 2 or 3.

Adam also shows in his video use of Magic Bullet Looks, an awesome suite of grading tools from Red Giant Software.

A quick video tutorial on how to set up User Defined Picture Style for video on the Canon 550D / T2i

It’s a simple flow of information from sensor through filter to card. This filter algorithm works out what information isn’t needed to be remembered for the video to still look okay. Its much like the way music is stripped of frequencies not audible to the human ear by the MP3 codec. It’s a simple case of space-saving for the card and buffer.

Now in most Canon cameras the filter algorithm is called the ‘picture style’. These are pre-set filters made by Canon, aiming to record the right information to make your video look its best, given you match what you’re shooting with what the picture style is called. So for example, if you’re shooting a landscape Canon think their ‘Landscape’ picture style would save the right mix of pixels to the card to make your video look just right.

Now these might be fine when you’re shooting photos in JPEG (they were originally designed to process photos not shot in RAW), but if you’re shooting video you’ll want to have control over what is saved to the card. Why? Well to properly colour correct your footage and get its looking its best you’ll need the right mix of properly exposed pixels to work with. So its really nice to know that Canon let you design and install on the camera some of your own picture styles. These can be designed in detail in the EOS utility and uploaded to the camera or quickly made in-camera by changing the Sharpness, Contrast, Colour Saturation and Colour Tone. You’ll also want to look at the White Balance Shift to take out any colour casts your lens or sensor may be making.

Now I know what you’re thinking, ‘Nick, what’s the best picture style for video?’ and the answer is it all depends on what you’re shooting, where you’re shooting and what you need to get out of the pictures. General consensus is that flatter and closer to a RAW photo the profile is, the more information is stored and can therefore be played with in post production. If you don’t want to put the time in messing with it in post you can always use a picture style that emulates your favourite film stock or effect. There’s loads of user defined styles out on the Internets so have a Google, load them up onto your camera and see what you like.

Personally I’d rather get a nice flat images so the camera records as much detail as possible in what will eventually be the highlights and shadows. This gives me more latitude to play with things in post and see where I want my image to sit in the grand scheme of detail and exposure. It’ll also give any colour correcting effects the maximum opportunity to keep detail in the picture.

Everything* in the visual arts can be traced back to the way the brain
processes the information given to it by the eyes and ears.

Roughly speaking the brain sees and decodes a continuous stream of information coming from the eyes. There is a point at which this decoded data can merge together. This can change given different lighting and motion conditions. Roughly speaking the point at which the brain cannot decode individual pictures and see them as distinct frames is about 25 individual shots per second (this is a massive subject area and would take hours of reading just to start to understand, so go with me on that, okay?).

Because of this, when television was developed it only needed to show a picture 25
times a second** to convince us simple humans that we’re seeing fluid
motion rather than a series of still images. Because of the way early televisions projected these
pictures they could only show half the frame at a time. So if you
slowed down a television picture enough you’ll see it only changes the top half and then bottom half of the frame at a time (progressive) or
alternately refreshes lots of horizontal lines (interlaced). The miss-matching
of these projection techniques is why computer monitors have that
scanning look when filmed. To make 25 full pictures a second the television actually shows 50 part-pictures.

Why is this important to us as DSLR filmmakers? Because these 50 part-
pictures dictate our shutter speeds. The choice of shutter speed should, for maximum
quality of picture, be a multiple of the frames-per-second you’re
shooting. So if you’re shooting 25fps you should have a shutterspeed
of 1/50, 50fps should have a shutterspeed of 1/100.

You should also take into account what you’re intending to do with the footage. If you’re going to slow the shot down in post production you should shoot at a multiple of the native shutter speed for your intended frame rate division. For example if you’re shooting at 50fps and intend to slow the shutter speed down by half to 25fps you should be shooting at double the native shutter speed, so 1/200.

So your shutterspeed is fixed, how are you going to control exposure? There’s two options, one is the aperture and the other is the ISO. One of the reasons for using a DSLR is to get that beautiful depth-of-field and aperture is your key to that,so you can’t really mess with it too much. Thats just ISO left and you can’t go down in ISO only up, which adds noise. What to do? What to do?

Well the best piece of kit to easily control exposure without having to mess with your shutterspeed, aperture or ISO is a variable neutral density filter, something that has been present inside of television cameras for sometime. This piece of glass attaches to the front of your lens like any other filter and rotates just like a polariser. Rotation of a variable ND produces reduction in light entering the lens and getting to the sensor leaving you clear to shoot side open and at your native shutterspeed. Its also infinity variable so when filming continuously from a bright to dark environment you can just rotate the variable ND as you go. They’re not cheap, and you need to make sure the one you buy doesn’t cast a colour across your picture or vignette your shot.

But if you’re serious about using your DSLR as a proper video camera you’ll be wanting this little bit of glass eventually.

*mostly, probably
**different frames per second are possible. These are all processed by
your brain slightly differently, making different cinematic effects.

Fact.

Fail to include incidental music and spot effects and your masterpiece is nothing but a shell of pretty pictures and (sometimes) shaky dialogue.

And thats where www.MobyGratis.com comes in to help. Its a free-to-register site that lets independent, non-profit film-makers and students download and use music from Moby in their projects. If you need to get a commercial licence then there’s a simple form to sort that out too.

Anyone making any kind of digital video will probably find this useful – MPEG Streamclip is a converting and transcoding tool which will let you work accross almost any video & audio format.

Already aware of VLC media player? This is the VLC of the conversion world. Even better, its totally free for both PC and Mac! Bravo to Squared5 the developers, big thanks for your work on this!

We have found through various tests that it is often better to take the native footage that comes from cameras such as Canon’s 7D and 550D and convert it from H.264 to a  less render intensive format such as XDCAM – this will allow you add certain effects in Final Cut Pro with minimal and in some cases no rendering time.