How I Made My Vid

February '07

Us

This vid, Us, is one I've been planning to make since August and I just didn't get round to it. There was so much source to collect, and then I had to make I Am/Lamb in a big rush of have to and then making I Am/Lamb destroyed my computer and I couldn't even get hold of Final Cut until my ecstatic post last month.

But now I have Final Cut Express and apart from one or two disasters it does seem to actually work, so yay. \o/

(I am coming to think that a mandatory part of making a vid is the whole thing getting hopelessly corrupted and you having to applejack up your OS whilst praying and banging your head on the ground: Oh, Vidding God, I offer up unto you my archived feedback folder and this selection of mp3s. I perform the Ritual of Rebuilding Launch Databases. Please! Take these piteous gifts. I ask only that you spare my children! )

There were several stages to making Us. My workflow was: Final Cut > Photoshop > After Effects.

Final Cut

I cut the video together in Final Cut Express and I think possibly the most interesting thing about was the speed changes that I used. When I listened to the song I had some very strong physical impressions, chiefly:

1. a sort of cuckoo clock —two mechanical figures sliding round a track and then jerking though the gongs, an elastic band stretching and then snapping back:

   [video, 868k]

2. A driving, rising motion that builds, is heavy and warped and pressured, and then bursts into a torrent of stimulus: a coming out, a crescendo, the city rising to the sea, your ears popping as you take off!

   [video, 1.5mb]

I thought of the music as having two distinct speeds, and three distinct parts: the fast piano that hoppped up and down, the, er, 60bpm? piano that to me was the gong, the clock —they were the beats and the jerky motions—and then the strings and the voice were the rising, lifting, smoother movements.

I cut on different parts of the music according to what was... this is the hard bit to explain. I cut on the voice of the music, which is not always the actual singing voice, but the narrative thread. So sometimes I was picking up the strings, like, I think, in the Atlantis rising sequence [00:00:32] I'm cutting on the strings there, but I did, once, I'd set up the sequence for the motion, slice it up on the 'gongs' [00:00:34, 00:00:35, 00:00:36] as well and put in additive dissolves (so the scene was dissolving into itself) to add light on those notes as the music rises.

I had little sets of three or four cuts of the same scene, with, say, a normal speed part, then a massively speeded up part, sometimes speeded up to 1000% or more, but mainly at about 400%, and then bringing the speed back down to normal or even slightly slowed time, like, a bit like an elastic band being stretched and then let go

After agonisingly huge stretches of wasted time some pissing around, I found it was easier to pull down the whole shot onto a drafting sequence, render it (I do a lot of boxing this way - three minute rounds on the bag while my bar is greening along), then cut away the chaff and make the speed changes and so forth. Then I'd drop that composed sequence of cuts onto my main sequence timeline on an overwrite. Making speed changes on the final timeline just fucks everything with a chainsaw. It's a disastrously awful way to work. I suspect that there is something I ought to be doing here, some way of isolating clips to perform processes on them, but I can't find it in the manual or online anywhere, and this method works for me.

[The Trek scenes, being so old, are cut amazingly slow and static. I found myself cutting out about 90% of each shot and just sticking the beginning and end of each cut together, like when they jump through the gate.]

+

Clip choices

I just don't know how to even begin verbalising this process. I can't do it, man. I can't talk about why I chose this clip or that clip. I can maybe talk about my methods though.

I have this set of criteria that footage must meet before it goes on the timeline. To my mind, there are three basic meanings to clips:

1. The Vid Meaning: what I'm saying with the footage.
2. The Canon Meaning: what the footage means in the context of the show.
3. The Context-Free Meaning: what the footage says visually (in the language of film, I guess).

So the trick is to find footage where the canonical and visual meanings support and underline the in-vid meaning. And not all meanings are created equal. An open window, for example, you might use to express opportunity, a chance at freedom, but if it's Angel's window and sunlight is streaming through it, the canon meaning of being imprisoned by the light of day might overpower that. In this vid there were some clips I wanted to use that looked great, but the canon associations were too strongly opposed - they drowned out my story, and some where the canon meaning was perfect but the actual context-free visuals were completely misleading, like droning white noise disappearing all voice into it. The idea is to have all three meanings strengthening each other, so the more you know about the canon or the visual language, the richer the vid becomes. In this way it is possible also to contradict yourself - to mislead or dissemble, and to pose questions, which makes it interesting to make, you know?

Once footage meets that criteria, I try it out - fit it for motion and light and so forth. In this vid I privileged meaning over motion, and although I did try to keep a throughline, it wasn't anywhere near as solid as the motion throughline in I Am/Lamb.

+

Bacon

I numbered the jpgs sequentially in Finder, pulled them into my project, selected them all and pulled them on to my drafting sequence as stills with a default duration of about 00:00:00;07 or something, and then used the push and slide transitions to create the movement. I added the motion blurring here too, to echo the comics sequence that comes earlier, and that was a terrible mistake because when I exported the video, and then again exported to stills, I fucked up the interlacing and got some seriously hammered stills that were just jaggedy lines. I repopulated the fields in most of the fucked frames in Photoshop but it was a hassle, and I actually ended up going through every single frame of the video layer of Ray and Fraser in Vault and blurring the odd fields by hand. Never again! Gah! In my next vid, I will work on interlacing... and add motion blurs last, in After Effects omg.

And that's it for Final Cut. I saved that video out as a huge FCE Quicktime movie.

[video, 22mb], ignore the garbage frames - that's corruption from the bad crash. You can see the terrible interlacing problems pretty clearly on there too.

I opened it in Quicktime and exported it to Image Sequence > jpg at 12 frames per second, which gave me 2579 jpgs at about 220k each. The first 75 frames were blank.

Photoshop

+

Then I spent quite a lot of time messing about with the jpgs in CS2 because I knew how I wanted it to look and it didn't look right. Although you can get a pretty decent pencil look with a bit of jiggering and a few layers, it's pretty hard to get something that you can then apply to different sources as an action or batch process and have it look ok. So, yes, a lot of fiddling before I got anywhere. I did go through and touch up the footage, deinterlace, remap the curves and all that for each different source, but I recorded what I did for each one and then ran batches on ranges of the jpgs. I essentially smoothed the noise, sharpened the edges, simplified the planes, and exaggerated the differences in tone. I'll go through a couple of actions now:

Used Batman Action:

1. An action I'd premade called [autolevels], which unpacks to Auto Levels, Auto Contrast, Auto Color
2. Curves > Understand curves here
3. Smart Sharpen (shadows)

4. On a new layer, I airbrushed out the Bat Signal and replaced it with a copyright sign that I skewed to sorta the angle of the original:
5. I saved that layer as a pattern and recorded Fill > Pattern > BatRights as the next step on the action, so all the rest of the jpgs in that range, the duration of that shot, all got stamped with the pattern and therefore airbrushed over with the sign. [I used pattern stamping to get rid of a few logos, and to alter various symbols and such, frex: the polaroids on Henry Jenkins.]
6. Select:All (command A) and Copy:Merged (Shift Command C) to New Layer (Command V)
7. The very last photo process in every action was to make it sketchy. I used a free trial of an amazing plugin for Photoshop called Alien Skin, and if you don't know it you should go and check out their plugins because they are fucking amazing. I wanted to be sure I had everything else sorted before I ran this step because it's a time limited trial: I had to get it all done before it ran out! I set up the parameters - the type of paper, the random seed, the length of the strokes etc, and saved those settings in the plugin dialog. Then I ran the filter (and saved that as an action called [pencillise])
8. Finally I played my premade action [saveout], which merged all layers and saved for web as a maximum quality jpeg.

Hmmm, looking at that result, that's actually better than my range, so I guess I must have tweaked these settings as I went through.

Once the action is set up, it's easy to go to File > Automate > Batch to bring up the batch dialog, and then choose Play > Set: Us, Action: Used Batman, Source: Opened Files, Destination: None. [I chose destination none instead of save and close because I wanted a new folder of optimised jpgs.]

I've uploaded about 1000 of these stills to this album: uscaps, in case anyone wants them for anything. I don't know why the other 1579 won't upload. It's a mystery.

+

After Effects

So now in After Effects I imported the JPEG sequence: Import > Footage: As JPeg Sequence and imported the video layer as well and set up a composition of 12fps, so when I pulled the JPeg sequence into the comp timeline, it was in time and everything.

I had planned to use auto trace on the luma of the video to generate a series of masks, copy these masks to all layers, and then animate the scribble function of After Effects to scribble the layers on to each other by filling in these masks sequentially. This is a relatively simple thing to set up because it's a straightforward process repeated many times, which is exactly what computers are best at facilitating. I spent quite a lot of time stuck in this idea because I had this idea that that was how to do it. *g* I had a plan. I planned it. I thought I was just doing it wrong and I kept changing the trace parameters and adjusting the stroke parameters and it was just wrong wrong wrong and I realised that the reason it was wrong is that it had no intent behind it, not really. I was tracing the shapes and scribbling in the video and all that but there was no... It was scribbling in stuff according to its composition (of shape and space and light or whatever) and I really wanted this sense of an author, of a person reaching in and drawing attention to this thing or that thing, or lingering over this thing, which is not always necessarily the brightest part of the picture, or the foreground of the picture so I was just like, ok, fuck it, I'll rotoscope it frame by frame.

A lot of the scribbling was rotoscoped, literally taking an eraser and removing part of the frame and then moving on to the next frame, each eraser operating as a static, single frame brush stroke. The paths, the scribbles that fill up and travel off I made by:

1. painting the whole stroke on the end state frame
2. extending that brush stroke layer's visible handle back to the start state frame
3. animating that stroke by putting a keyframe on the first frame of End > 0% and a keyframe on the last frame of End > 100%.

That was the most basic method. Sometimes I keyframed in speed changes, and I keyframed a few different values for angle or opacity on the relatively static shots, like the Joker sequence [00:01:52]. I ended up with 707 different brush strokes on the pencils comp and then handfuls of brush strokes on different video layers or sequences (the Krycek sequence [00:01:25] has the pencils painting on to the video, frex).

Animation

I cannot go through every single animation because it would take me longer than it took to make the vid and we'd all die of boredom. Also, really, I'm using the same principles over and over, in different combinations and with varying levels of complexity. It's better to grok the underlying principles than for me to write out a load of button presses, da?

Most of the transition animations were composed in my work composition, then precomposed and dropped back into that composition as a layer. I do it this way because I am lazy and cannot be bothered lining everything up so I just work from end frame to start frame, precompose, and then change the composition duration in the settings window (command K). I'll go through a couple of animations now.

Matrix Upon Sea [00:00:38]

I spent a while thinking about this one. There are numerous tutorials on the net on how to create a very good copy of the falling lines of code in Photoshop, which I did, and I think the actual effect was made in After Effects anyway, using a pro tool called Text Matrix, which I don't have. But, you know, I made this thing - I downloaded the Matrix Code font and followed these instructions and it was pretty perfect and I was just like, damn, what is the point of this again? I might as well just clip from the film. Anyway, then I remembered that I really wanted to evoke the Matrix, not perfectly replicate it, d'oh. And I wanted to explore my themes of growth and expansion more thoroughly so I scrapped all that noise and made my own. *g*

All I did here was really combine the process I had used to make the scratches in I Am/Lamb and the Photoshop tutorial I'd read about the Matrix. But I'll write it out anyway:

1. I went to Layer > New > Solid (command Y), and made a solid with the dimensions and duration of my composition
2. Then Effects > Noise and Grain > Fractal Noise
3. Then I futzed with the fractals until I got a decent linear string with a low complexity, and then animated its evolution and offset over the duration of the transition. To get your head round manipulating fractals, I suggest doing exactly this: make a solid, add the fractal noise, and then change the settings to:
   • Fractal Type: Basic
   • Noise Type: Block
   • Contrast: 100
   • Brightness: 0
   • Overflow: Wrap Back
   • Complexity: 1
   • Blending Mode: Multiply
   Once you've got your solid set up with these properties, move up through complexities. This should make fractals suddenly clear to you.
4. I set the blend mode to add [see: the subtractive colour model]
5. I generated an Audio Amplitude Keyframe Layer and with a simple expression, keyed the evolution of the fractals to the strings in the music.
6. Then I duplicated this solid about six times. I scaled up some of them, to create an impression of depth, added Find Edges to some, and blended all the solids together using multiple blending modes. I also animated the opacities.
7. I made a text layer and wrote, in Matrix Code font: "they made a city out of us", and some of the John/Rodney story I was writing, and quickly animated that raining down the screen (horizontal type, effects presets> raining characters in, I suspect) and folded that into the fractal layers
8. I went to Layer > New > Adjustment Layer, and adjusted everything to neon green
9. I made a new solid, on the top of all these, and generated a new fractal - some basic smeary clouds with quite a high contrast. I auto-traced this solid on the luma, and copied the masks this process generated. Then I deleted that layer.
10. Then, back in my main composition, I pasted these masks onto the matrix composition-as-layer, put a pretty high value on "Mask Feather", and subtracted the masked areas to create texture and depth.
11. Oh! And I set an opacity keyframe of 100% on the cut between Atlantis and the Matrix, and faded it out smoothly on both sides of the transition.

I used, font wise, Barcode MK, Matrix Code, Mulder's Handwriting, and Courier. I did initially have a sort of —the comp was called 'evolution of type' — thing going on over the top, with the green courier (which was what our first PC, an Amstrad, had when I played the HitchHiker's Guide to the Galaxy text adventure, which I was hopeless at because I had not read the book. Because I was six. ) to black courier to times new roman to verdana and so forth, running through this story I was writing at the time, which kept showing up different formats, like IRC to LJTalk and all that stuff

In the end I seriously cut back the amount of text because it was too much. I had these elements to kind of shoehorn in: the trees, the books, the fic, the scribbling voice; the screen was too busy. So there are only a few parts visible and I only had the green 12px courier font in the LJTalk formatting with the timestamps actually coming through. I animated that using "Word Processor", which is a basic slider expression preset. On the Gandalf text burst I keyframed in the position property and scale property of a mask that rose up to reveal the text and then keyframe-animated the text bursting out from the mask by easy-ease scaling up the text and easy-ease fading out the text with a keyframed tracking expansion. To keep all the movement together I animated the values on one property and copy-pasted the keyframes from it to all the other animated properties. This is a really efficient way to make apt and unique text animations and I will always use this method in future.

Polaroid Scatter [00:00:31]

A couple of people have expressed interest in this animation and it was seriously a piece of piss so let's go through it:

1. I made a simple black and white polaroid-like frame in Photoshop.

2. In the action for the defined range I recorded a pattern stamp of this frame.
3. I isolated this range of jpgs and duplicated them into their own composition, above the main comp.
4. I masked out the black surround, and copied this mask to every jpg layer.
5. I sequenced the jpgs and then, with them all selected, extended their duration so they all ended on the same frame (but began in sequence, still)
6. I stepped through the sequence and rotated and moved each jpg layer to make them look like they were thrown on the floor.
7. I added this composition to the "pencils" composition, and then stepped right back into the main comp and erased the inside of a few polaroid frames to reveal the video.

Er, are we done? I think I'm done typing, dudes. OMFG. Did I miss out anything really important? Is anyone still alive?

eta: Oh, the trees. I made the trees in Context Free using a very simple grammar of repeating shapes on a slight rotation and a slight scale reduction, or (circle x 2) (r 45) x 0.99 .