Tuesday, April 3, 2012

The illusionist: research and development(an awful lot of)

Christ, don't read this unless you must.




Research/development

The magician

Project output: a promo video sequence for an entertainer to impress his potential clients. The length of the clip should be anywhere between 60 seconds and 2:30.






Project background:
I started working with Radek in December last year when he performed on RGUnion's Talent Search.
Soon I realised that working with someone who focuses on illusions and sleight of hand is much more difficult than filming an event or a sport performer. I found it difficult to adjust the display frame at the beginning as I would not know what to expect: while filming Zahed Cheshty for Streetsport it was pretty straight forward to me: all I had to do was set the camera right, press the record button and occasionally tilt the tripod in one way or another.

This time it was different as I would not know what to expect. In a performance of this sort, the close-ups  are the key to a succesful video - even if the show is spectacular but the details are missing it renders the video coverage useless. I would not know where to expect a coin/or a card to appear in the next frame.

Unlike some of my previous projects it was not something I could direct and produce 'on the go' and hope for the best results.



Research:

My research mostly is centered around promo videos of similar sort found on youtube.com and other video websites.
There is a variety of groups who specialize in similar projects (e.g theory 11's channel). Their objective mostly is to promote accesories throughout the performes and advanced visual techniques. In my case however I am developing a sequence which would focus on promoting a person instead of a product.

I did also see some full feature productions like The Illusionist (2006) and The Prestige (2006) to get a grasp of 'being a magician' and what values are key to a magician's performance. Of course both of them are fiction or fiction-based productions and are placed in previous centuries but they describe quite accurately the process of the illusionist designing the trick.

What I learnt from the first filming session with Radek was that someone who is into the illusions and sleight of hands perceives things completely different. I put together quite nice looking clip of his performance and in a message back I got a list of all the moments that would have to be cut out. "Why" I asked.
"Because all of these explain how I actually did the trick".

The key to a magicians performance is making sure that the eyes of the audience follow exactly where he wants them to be. Again this was quite difficult - after several sessions I could get a grasp of what actually happens in front of the camera, but still while filming I had to pretend I am just an ordinary viewer who does not have a clue.


What I got out of the theory 11 channel was that a trick looks best if it is filmed from several angles (best simultaneously, but since I am a "one man's army"  I had to film the same trick several times and hoping that one attempt would be as succesful as the next one: the key to putting this sort of footage together is to get the sequences as similar as possible.

What I had to make sure of was that whatever is happening in front of the camera it has to stay in the frame: if even for a brief moment Radek's hand would disappear from the frame someone could accuse him of "forgery" which would completely ruin his reputation (at least among the illusionists).




Development:
Time remapping.  I realized that the best way to 'mesmerize' a viewer was to play something even slower. This is a good answer to the popular 'oh you just did this quickly, that's all". If even in slow-motion the audience can't tell what is going on is even better.

Color corrections: In order to raise the quality level of the project and add some dramatic/professional look I decided to use some color correction software (mainly Red Giant suite).


Lens choice: This was a bit tricky. Normally I shoot just about everything using my standard 50mm lens which gives pretty neutral output. I did have an opportunity to test Canon's 70-200mm (2.8 aperture) which is a considerably more expensive piece of equipment.
Although the quality of the picture was spectacular it was close to impossible to maintain the stability of the image. This was due to two reasons:
1. Shakes. While zoomed in on 200mm (which is *not that much* compared to 500m for instance) every even slight camera shake is multiplied on the image. So for instance if the camera experiences a rapid movement to the right by one inch, the footage will experience approx movement by 3 inches.
2. It's heavy! The weight of my 50mm lens is approximately somewhere around 100g. This giant pipe is somewhere between 2 and 3 kg. If you add the camera body to this (another kilo) it comes down to quite something (especially if one intends to shoot for several hours continuously). The funny thing for me to notice was  that the lens had an actual tripod screw, so unlike the other lenses I have seen while using tripod/shoulder rig or any other stabilizng gear one should screw it on using the lens instead of the camera body. This is due to the height and size of it.
3. Zoom + Focus. As I am accustomed to using prime lenses, havin a zoom lens is quite a challenge. First off there are two rings to control. In theory, the zoom ring if shooting default footage is to help with a better frame and should be left in this position for the entire sequence. It did happen to me several times that while trying to adjust the focus I would often use the zoom ring instead!
4. Minimal focal distance This was a tricky again. Even though this is a zoom lens it requires somewhere around 2metres distance for the subject to come into focus. This is quite difficult especially that even on the full zoom-in 2 metres is pretty far away if one needs for instance a take with character's eye filling half of the frame.

After giving 70-200 some trials I figured I would still be better off with the prime lens (at least for the dynamic footage). On the other hand, the tripod footage is superb.

To sum up: combine 70-200 tripod footage with 50mm handheld footage



Case studies


Solo by Michael James (official) source:youtube.com - http://www.youtube.com/watch?v=H5sKEveJB1U


This promotes a product which is a set of guideleines on how to perform some tricks the package features.
Opens with a sophisticated animated logo not sure if Radek has one like this.
The video features a variety of close-ups. Opens with a sequence on a luxurious watch, then centers on a tie. The next scene features a corporate skyscraper. These symbols are associated with a certain social status. The company is imposing a social image of luxury by this much like commercials always include a message 'by purchasing this product you would buy into a social standar'. However unlikely this is, works most of the time.


To follow the technical point of view, it operates heavily on bokeh sequences (very low aperture, again the prime lens is the key).  (0:45) - a good example of interrupted/cut into sequence. This is contrary to slow-motion footage and gives a chance of speeding up the process a bit. This should be mostly used on the 'filler' footage rather than tricks.
When the text bits appear (e.g 0:48) they are cast over a semi-transparent solid so while being able to see what is going behind, the viewer can still read the information. This sort of 'two in one' again allows to sequeeze more information in a shorter period of time.


STRUCTURE(!):
Typically for a movie trailer the footage consists of two parts and are subsequently divided by the background music.
Part one (approx until 0:54) has a pretty sublime soundtrack on a slow-pace. This is the build-up:

This part introduces the main (and only) character as well as some key information on what product it advertises.

Part one (0:55 onwards) 'wakes you up' : from this moment the sequences are way more rapid, the music changes into more dramatic/epic (in a movie trailer this part would feature an epic fight or gunshot sequence or something). These sequences are brief and include some rapid zoom-ins. There is an easy way to achieve this. While working with digital SLRs one can shoot the full HD (1080) footage. Then, while working in 720 projects in Adobe's Premiere Pro/ Final Cut, simpy work on scaled files (720p is 67% of 1080) only to switch back to 100% for a brief moment.
At 1:07 it slows down briefly -slow-motion footage-

Further visual methods include:
the usage of a lens flare (1:11)
lighting effects (1:24)
film burns/light leaks (0:36, 0:46, 0:55, 0:59)




MONARCH PLAYING CARDS - new preview! source:youtube.com - http://www.youtube.com/watch?v=Fpmdh3Sk84U&list=UUl1WwbVpKUdQerivZpUxZLA&index=1&feature=plcp

This is just to follow the previous case study and emphasize the trailer stylization. Putting the title aside the sequence does not suggest until about 0:47 that it is actually advertising a set of cards.

The clip is structured like an action-movie trailer again featuring a lot of rapid scenes of a busy city. The dynamic music again speeds up in the second part, as mentioned in the example above. Unlike the previous production it features some more close-ups on the deck of cards as after all this is the advertised product.

The best way to sum it up is to quote one of the comments on the video: 'it's hard to believe this is just a playing card commercial..".











Shooting session 2:
22/03/2012 at Union Plaza

Contrary to our first session at the Social Sciences building of RGU, this was shot outdoors.

I chose the Union Plaza area because it is one of the most urban/big city areas of Aberdeen.  And again we want to promote an image of modern luxury, except our budgets would not allow us to travel too far.
Ironic.

This time I had another lens to test, Canon 85mm f1.8.
Contrary to what the price might suggest (approx 260, compared to ‘mere’ 80 pounds for 50mm 1.8), I could not tell any difference except the focal length of course.

Since during our first session we covered most of the ‘trick side’, this session was meant to provide us with the filler footage (see in Case studies section, mainly the first parts of analysis) for the build up part.

As we discovered the area though, we happened to come up with some interesting ideas for using the urban landscape and the Union Plaza’s architecture to our benefit. The low-light square, strong red light of the entrance wall and reflective surfaces of the building walls were among the elements we took advantage of.
Still, this session was mainly about getting ‘extra’ footage and material that looks nice by itself.

What I learnt in the process:

The repulsive nature of Aberdeen’s weather has not helped us much.
As the performance took place outside, it was difficult for Radek to do much with the cards due to the strong wind.
Also as the temperature surprisingly dropped to 2 degrees not long after our arrival, again cold hands get in the way of executing sleight of hand smoothly.



For session 3:

In order to get smooth slow-motion I need to film in fixed conditions first.

(from Philip Bloom’s blog):



I have used twixtor myself for a few projects but I have never had terrific success with it. One of the best filmmakers I know who uses Twixtor really well is my friend Salomon Ligthelm, whom I met when he lived in Dubai back in January last year. He put up a new video a couple of days ago, Gravity, and he was kind enough to write this guest blog post on how to get the best out of it.
Twixtor will not give you as good results as a dedicated high speed camera of course. They do that job incredibly well. You need to light a lot to make them work but as you will read below you need to do the same with twixtor to get the high shutter speeds needed. But for those whose budget means renting a Phantom Flex or something similar is out of the question Salomon guides you through the best ways to make the most out this powerful, but tricky plug-in. 

Let me quickly explain how Twixtor works: It takes your footage [generally your transcoded 50fps, or 60fps footage and "guesses/interpolates" your missing/in-between frames
I guess with every project the approach is somewhat different. When Twixtor becomes a part of my workflow I need to become increasingly more aware of the way I am shooting so that I have workable footage in post.
I wanted to shoot something different to all the other Twixtor stuff I've been seeing - which have all been shot outside. I wanted to try shooting super slomo using the Canon 7D and Twixtor in interiors.
Here are a few key things to consider when working with Twixtor:
Shooting Workflow:
1. Shoot at a fast shutter: anything 1/2000 - 1/4000s. [so your lighting needs to be good]
2. Shoot against a plain/solid colored background – sky/wall etc etc.
3. Shoot at the highest point of action. The law of nature is that what must goes up, must come down. Its the movement at the top of an arc/trajectory that works best for Twixtor because that’s when things generally move the slowest, more frames of the relevant action for Twixtor to work with.
4. When you shoot really close/tight, slow the action down manually ie – move slower and fake slow motion [all the tight sequences in 'Gravity' was done this way and then slowed down a little more with the tools]. Its easier to get good results with Twixtor when the action is shot from a distance, because of the fact that the action will happen across more frames – ie – Imagine moving your arms up and down – a distance 50 centimeters right in front of the camera – the camera will only catch SOME of that action as your arms pass by the front of the lens, for maybe 1-3 frames. However if you move away from the camera and shoot say 10-15 meters away from the lens, you would catch the ENTIRE action across say 7-10 frames – giving Twixtor more information to work with
5. Shoot at 50 or 60fps. When you shoot high-speed you give Twixtor more info to work with – it is a shame that you have to forgo resolution
6. Shoot as much as you can – try to get the action from different angles – reshoot stuff if wasn’t working, or if your settings weren’t correct. SHOOT MUCH!
Post Workflow:
1. Use Twixtor in After Effects – for some reason Twixtor works best in After Effects – for me at least. Probably because After Effects’ ability to manipulate time is so effective.
So here’s what I do in a typical scenario.
- Edit the entire piece in an NLE [Non-linear Editor - like Final Cut] and then I’d
leave sections in the timeline open, or maybe reference clips that I want to
work with eventually in After Effects using Twixtor – you can’t do too much in
FCP at this stage because we’re going to be playing with time-changes – which
can screw up things if you don’t work carefully]
- Move the whole piece over to After Effects via an XML script called Popcorn
Island FCPtoAE. You need to export an XML from FCP by going to File >
Export > XML [ for FORMAT click Apple XML Interchange Format, Version 5]
- Open AE and go to scripts – look for FCPtoAE and then pull up the exported
XML
- Now click on the clip you want to use Twixtor on:  from the MENU go to LAYER
> TIME > ENABLE TIME REMAPPING. This allows you to extend a clip our for
longer – because when you start slowing stuff down, the clip needs to become
longer, which you can now easily do by enabling time remapping – just pull out
the edges using the drag function on the clip
2. Check that your Twixtor settings match your footage ie – Make sure your frame rate is correct when you set up Twixtor’s settings. If its not correct then Twixtor will stutter across your footage
- I typically shoot and edit in 23.976fps because I like the film aesthetic.
- I generally find that manipulating the image to a slower speed – say anything
between 2-5% yields better results than trying to get something workable at
30%. I don’t know why that is, but I reckon it has something to do with the fact
that with many more frames created it uses those created frames to reference
from in order to create more ‘interpolated/guessed’ frames.
3. Consider using an aspect matte to hide warping. I often use an aspect ratio of 2.4:1 in my work as a lot of my warping happens at the edges of my shots, thus a matte will hide the really bad sections. One instance of this was a shot in the short of mine called [Fo'tis'ma], where my good friend Franzwa is taking off from a ledge [Its the tight - slightly angled down shot at 34s]. The warping on his hand was REALLY bad – especially because the shot was rather tight and he was coming down at such a pace. It was because of that shot not working at 16:9 that I decided to make the whole piece 2.4:1 [IMAGE 1 & 2]
4. Avoid warping by editing well. Many times Twixtor works for only a few frames – if that’s the case I’ll only use it for a few frames and will cut to the next image – to avoid things looking crappy [obviously this must be done tastefully]. Don’t hang on too long on a shot that was good to start with but warped into mush at the end – EDIT WELL.”


The difficult part in owning 5d Mark II (other than being amazing) is the fact that it cannot go above 30 frames per second.
For this reason, I had to step down several models and get Canon d550 which will allow this except I will for-go the resolution (720 contrary to the crisp 1080 5d provides).
The drawback of this is that I won’t be able to make any amendments to the composition afterwards – while I shoot at 1080 I still work in 720 project files due to two reasons. One being that 1080 is not yet widely accepted and most of the internet services will either ask for 720 pix or “downgrade” it to this.
The second reason is that by placing 1080 footage onto 720 project it allows me to reposition the footage the way I like best  or allow zoom-ins without the loss of quality,

No comments:

Post a Comment