Thoughts on RED footage

I don’t think there’s a camera brand that polarises filmmakers as much as RED. Some people are blown away, signing up for a life on the forums of REDuser.net, while others deride it. And I’ve heard from at least three people that despite the hype, watching the footage back in post was extremely disappointing — the blacks were washed out, the colours were dull, etc etc. And after grading and transcoding 520GB of R3D footage last week, I finally feel entitled to have an opinion on the issue. I’ve come across some strange anomalies when viewing the footage in a variety of programs, and it’s difficult to know what to trust. I opened everything up in REDCINE-X and graded until I was satisfied with the result. But upon exporting a 2K Prores 422 HQ version of the clip and looking at it in Quicktime X, it was indeed a disappointment. Despite crunching the blacks in REDCINE, everything looked washed out and dull, with significant noise in shadows. However, calling up the same ProRes clip in Final Cut 6 yielded far better results, much closer to what I’d previously graded in REDCINE. Finally, I placed the clip onto an FCP timeline and exported to h.264 640×360 — this h.264 version played perfectly within Quicktime X, with deeper blacks and much better contrast. Yet all I’d done was pass the footage through FCP, without any additional filters or processing. I can’t explain the discrepancy, but my hunch is that Quicktime handles 10-bit images differently to REDCINE or FCP. The final h.264 transcoding knocked the footage back down to 8-bit, possibly making it easier to display in Quicktime. It’s a fairly half-assed theory, I know.
In any case, the footage is looking good in FCP, and from there I can export back to quicktime and be confident of the results. So yes, any idiot with a Mac Pro in his bedroom can make RED footage look “good”. I’m not surprised. Now.. the more interesting question is “was it worth it?” — why couldn’t we have shot on a DSR-450, or an HVR-Z7, or a 5D? What benefit do we have from working with RED footage? Well, the knee-jerk reaction is to say “The DOF man, it’s the RED’s DOF that’s awesome it’s like that kind of DOF you get when you’re shooting on film or something man it’s just awesome” and so on. But.. let’s take DOF out of the equation for now. Let’s talk resolution. On one side of the fence, I realise that we don’t need 4K. Further, I realise that the RED One cannot resolve 4K anyway. But on the other side of the fence, I think that standard definition PAL video stands out like a sore thumb these days. The aliasing can be quite severe, especially in wide shots, and for a large-scale screening, you can definitely see the difference. Online, PAL video is still alright, but SD is still a significant limitation. That eliminates the DSR-450. But why not shoot on a Z7 or 5D? In terms of resolution, HDV is getting there (1440×1080), and the 5D delivers full 1920×1080 HD. Let’s assume we’re mastering in full HD — both HDV and 5D will get us over the line. The HDV footage would be undersampled, and the 5D footage would be sampled at just the right size.. compared to the RED’s quadruple supersample. Ultimately, after transcoding everything to HD ProRes, the difference in resolution would be fairly negligible. You’d see some benefit to supersampling at 4K and downconverting to HD.. but not enough for “regular audience people” to swoon and fall out of their chairs.
But now let’s talk compression. Here’s where I think the RED footage really shines. I don’t know anything about the inner workings of “REDCODE”, but it’s flexible and robust from what I’ve seen. Footage can be pushed around. A lot. And compression artifacts are minimal or absent. There’s some banding in deep shadows, but nothing that a bit of crunching can’t fix, and the 12-bit tonal range is really impressive (I say this because I’m used to DVCAM). Now.. fans of the 5D are going to be telling me that I shouldn’t *need* to push the footage around in post. If I was any good with a camera, I’d light it and shoot it the way I wanted it, deepen the blacks slightly in post and that would be it. Today I ran some tests on some 5D mkII footage (I managed to get my hands on a 5D for the first time last week), and the results scare me a little. When viewing at full 1080p, it’s all quite decent — you’re only going to find artifacts if you’re specifically looking for them. *But* as soon as I desaturated the image in FCP, it all started to fall apart. All I did was take the colour away, and suddenly my image was composed of obvious blocks and squiggly mess, with serious banding in the shadow tones. The compression damage quickly worsened as I boosted the contrast, which made the banding ridiculously obvious. This all scares me because my RED shoot is incomplete. We still need to shoot the first scene, on a 5D, in black and white. And the results of my testing do not bode well. We’d be fine if we down-rezed to PAL, or maybe 720p… but at 1920×1080, there’s a massive difference between my RED footage and 5D footage.. much more of a difference than there would be if the film was left in colour. I’m hoping that if I shot the 5D footage in black and white to begin with, the h.264 codec would focus all of its attention on making the monochrome footage look good, but I have to test this to be sure. Otherwise it’s back to the drawing board.
So anyway. I’m extremely glad we didn’t decide to shoot the entirety of our PP2 film on a 5D with the intention of desaturating it in post. That would’ve been a nasty and fairly devastating surprise. Regardless of depth of field or “4K” resolution, I think the core reason why we forked out extra to shoot on RED was the sheer malleability of the footage and the compression codec. REDCODE is a fairly phenomenal thing.

Shots of the day – RED

Gamma Problems

I don’t think there’s any answer. I can’t trust any monitor. I can’t trust calibrated monitors because your average TV or computer screen is not calibrated. And I can’t trust uncalibrated monitors because.. they’re not calibrated. In snow leopard, the new mac OS, gamma is automatically set at 2.2, which matches the gamma setting on PCs. Of course, not everyone has Snow Leopard yet. I’m currently viewing my uploaded videos on an older operating system, presumably with the old gamma setting of 1.8, and I find all of my videos to be washed out… it’s quite scary. But then, I’m viewing this on some HP monitor that seems to be washing out a lot of things.. so maybe there’s no problem.

It’s just frustrating that colour/gamma settings are so relative. When you buy a TV, they pump up the saturation and contrast so that everything looks awesome. And so if I go into Final Cut and pump up the saturation and contrast of my footage, then export it to DVD and show it on a TV, my footage looks fried.. way too saturated, way too contrasty. The alternative is that you grade your footage *knowing* that the TV settings will make it look awesome, but TVs are all set up differently, as are computer monitors, and then if your footage looks bland, that *is* your fault because you could have prevented it from happening.

Apparently, newer codecs such as h.264, have adaptive gamma capabilities — quicktime and other players can read the footage and automatically adjust the gamma space so that it’s corrected for your operating system (not sure if they can adjust it according to your monitor setup though). That’s all well and good for online videos shown in quicktime, but exporting to other web codecs, or exporting for DVD or projection is another story. I just.. don’t know who to trust.. what exactly I should do when grading footage for exporting to multiple platforms.. I guess ideally, you’d grade the footage differently for each platform, but settings are so relative that I don’t know if there’s a “right” choice, even on a platform-to-platform basis. blah.

It’s turning out all wrong

So, the core of our assignments, the core of PP1 itself is the ABC Pool website. And despite this, I really know very little about how it operates and what its technical limitations are. I used it briefly last semester to upload my Machinima, but that’s about it.

—Until I visited yesterday and attempted to upload a video. Now.. the first thing that strikes me as I look at this site is “wow, what is this horrible compression codec they’re using on all our videos?” And then it all came rushing back to me. Last semester I’d painstakingly exported my Machinima project at various bitrates of h.264 compression, trying to maximise quality while keeping under the file size limit (75MB). I ended up with a beautifully compressed export at around 56MB. It was high bitrate, 25p, 16:9, 640×360 pixels downsampled from 720p (I had to massively compress the sound quality in order to compensate). And after patiently uploading to Pool, I could finally view the finished product online… “Oh god. wtf happened Pool what have you done omg” was pretty much what went through my head. That is one vicious codec they’re using. Vicious. I’m guessing it’s flash encoding.. with keyframes every second or so and a low bitrate.. it handles still shots really nicely, but freaks out at the slightest of motions. And.. don’t even think about using wide shots.. the blocking artifacts prevent you from seeing much of anything. Also, the player only does 4:3, so your wonderful widescreen creation will be letterboxed, and the final resolution most likely sacrificed as a result. bleh.

I’m not a web designer or a video technician. There are probably very good reasons why the Pool has such an evil compression scheme that I do not understand. But regardless, my initial, gut reaction is a user of ABC Pool was still to shudder and recoil in the face of a codec that has maliciously raped the visual quality of my online masterpiece (or whatever you want to call it). So.. as a content creator on Pool, what can I do to avoid this horribly compressed look on all of my video works? According to the FAQ, Pool accepts MP4, MOV, and WMV. MP4 and MOV files are both to be compressed using the h.264 codec, which is what I’ve already done for my machinima. h.264 doesn’t fare well when Pool recompresses it.. I know that for certain. What I haven’t tried is uploading a WMV.. maybe WMV files can be converted to flash without as much messiness. The other possibility is that the Pool compression will work better with a 4:3 video.. though I think that’s a long shot. I guess I’ll try both. If I come up with a useful method, that could be a good thing to spread around Pool.. but I’m probably being overly optimistic. Perhaps no export format will go unscathed in Pool

http://pool.org.au/video/kenneth_stirling/the_garden_of_gethsemane

After trawling through a few pages of video uploads, I noticed that the above video actually fared well against Pool’s compression (have a look at my old machinima upload, for an example of footage that *does not* fare well). Of course, this is most likely due to the high contrast and complete lack of colour in the video.. there’s a lot less information to compress, especially as large portions of the frame are flat white or flat black, and motion is fairly minimal. So.. it looks like the main problem with Pool’s video compression is a low data rate. it seems to handle footage reasonably well if there’s very little visual complexity. hmm. Doesn’t sound like a stunning breakthrough.. I can’t really tell people on Pool “Hey man, awesome video. I noticed, however, that it’s quite compressed and there are lots of artifacts — perhaps for next time, you could reduce the visual complexity of your compositions? Cheers”
No.. doesn’t sound promising.

I may as well try uploading a 4:3 video work as a WMV and see what happens, but I think that the low data rate on Pool is the primary barrier, and short of reducing complexity in the frame, I’m not sure there’s much else that will help.

some kind of workflow

Why has it been so quiet here as of late?

Well.. I’ve been away. And now I’m back, but I’m in post-production hell. 6 episodes, 22 minutes each. 15 days of unlabelled HDV footage from two cameras, only one of which has usable audio. Most shots are slated, but there are various mystery-clips that have no slate, and many others have a flash of slate for three frames in the top-left corner of the shot, out of focus with motion blur.. usually you can guess what take it is though.

Post-production hell has its benefits, however. The job will cover the cost of my new 4-core Mac Pro, which I sort’ve need anyway in order to cut a whole project in HDV (I’d need at least 2TB of space to transcode all the footage to ProRes, which would’ve been necessary for it to run smoothly on my old 2006 iMac).

I realise, I’m not an editor.. and a 4-core Mac Pro with 8gig of ram isn’t really necessary unless I properly want to be an editor. But one implication is that if we shoot anything on RED this year, we’ll have an intel-mac to run REDALERT and REDCINE-X etc on. (the RMIT computers are good, but pre-intel.. which is a pain). The Panavision money would cover two days of shooting on RED (it’s 900 per day, but another 5-600 for lenses, adding neatly up to around $3000).

I’m thinking that in post, the best option for RED shooting seems to be using the Quicktime proxies which are part of the larger 4K shots. I’m reasonably sure that this requires no processing, just the installation of a plugin in FCP. Using proxies would be faster initally as there’s no process of transcoding everything to ProRes, and hence you also keep all the flexibility of RED’s ‘RAW’ codec.. would be pretty good for colour grading, I imagine. Especially since any colour grading of the 4K source files automatically alters the appearance of the proxies to match.. And (using CRIMSON, I think) at the end of your edit, you can link your proxies back to the larger source files and finish the project in HD.

So yes. There are benefits to this job I’m doing. But I’m eager to get back into production, it’s been too long.. and it’s frustrating having to spend every day doing at least 8 hours of editing when I could be applying to assist in productions or whatever.

Over the next few weeks, before uni starts, I need to figure several things out, in terms of this final RMIT film I’ll hopefully be making.

I need to visit Lemac, and possibly other places that offer telecine services, and ask for quotes. I also want to know what formats they can scan in.. ideally, an HD or 2K scan of our 16mm footage would be great, but the cost might completely destroy us, I’m not sure. It just seems a waste to shoot on 16mm and then end up with footage on digibeta..

Once I’ve figured out the telecine options, I can begin to figure out whether this project would benefit from being shot on multiple formats.. namely 16mm and RED. Obviously, you’d need a stylistic justification for using multiple formats as they’ll look significantly different. But assuming we found a justification, we could effectively afford two free days of shooting on RED, and we’d be covered for if we ran out of film stock. It’d allow for a larger-scale shoot, since there’s no way we could afford to shoot for more than a couple of days on 16mm.. if we shot on RED as well that’d be 4 days.. that’s a lot more production time than I’m used to on RMIT films.

Then, assuming that multiple formats sound like a good idea, there’s the question of framerates.. do we shoot everything in 24p, or shoot the RED at 25p and mess around with the 16mm footage in post..
And after that there’s the question of aspect ratio.. although that’s fairly easy.. 1.78:1 is most likely the best option, now that I’m sure the SRII can shoot in 1.66:1..

Many things to consider. But first.. over 100 more hours of editing.

Blowup

“Several other factors, such as subject matter, movement, and the distance of the subject from the camera, also influence when a given defocus becomes noticeable.
The area within the depth of field appears sharp, whilst the areas in front of and beyond the depth of field appear blurry.
For a 35 mm motion picture, the image area on the negative is roughly 22 mm by 16 mm (0.87 in by 0.63 in). The limit of tolerable error is usually set at 0.05 mm (0.002 in) diameter. For 16 mm film, where the image area is smaller, the tolerance is stricter, 0.025 mm (0.001 in). Standard depth-of-field tables are constructed on this basis, although generally 35 mm productions set it at 0.025 mm (0.001 in). Note that the acceptable circle of confusion values for these formats are different because of the relative amount of magnification each format will need in order to be projected on a full-sized movie screen.”
http://en.wikipedia.org/wiki/Depth_of_field

So.. if your final output is 35mm film projection in cinemas or whatever, and the apparent focus of 16mm footage becomes dubious when blown up to 35mm size.. does this then cancel out the effects of using wider 16mm lenses with deeper perceived focus? I mean.. you’re using longer equivalent lenses when shooting on 35mm, but if the tolerance, the margin of error is more forgiving because there’s no blowing up of the image, which format ends up being less forgiving in terms of perceived focus when outputting to 35mm film?

This is where I rush to my nearest film lab, beg for their 16mm short takes, and start testing in my backyard on an ancient, makeshift film camera.. I wish I was old-school.

Colour grading tendencies

You’re in Final Cut Pro. You’ve been cutting a doco, concentrating on how the story flows and how the audio cuts together. You’ve been editing for weeks and become so familiar with your footage that you can’t conceive of it looking any different than it does right now. But somehow, it’s now time to colour grade.

Familiarity is the enemy, I’d argue. It’s easy to say “That filter looks weird, I don’t like it” when you’re so used to how the footage originally looked. You’ve got to keep remembering that the audience only sees your film once, and will probably embrace even a radically altered colour grade, as long as its consistent. They don’t know what it used to look like, so it’s not nearly as jarring.

This is a bad example, but consider “300″ (naked roaring spartans killing middle eastern people for two hours). It’s got an intense colour grade, some kind of hyperreal super-sepia look. But after the first few minutes, you embrace it, and the film plays out as normal, despite the fact that the entire film is practically monochromatic.

So that’s what I’m getting at. Anyway, let’s continue. We’re in Final Cut, and we’ve got a vast array of filters to use.. lists upon lists.. where do you start?

I generally throw a “proc amp” on any given clip, followed by a 3-way colour corrector, and then another proc amp. That’s my basic set of filters to mess with. The proc amp controls the blackness of your blacks, the whiteness of your whites, and the saturation of the whole image. You can use the 3-way colour corrector for the same things, but I think the 3 channels are broader on the colour corrector (ie when you blacken the blacks, it darkens a lot of your image, whereas the proc amp only effects the darkest regions).

So you crunch the blacks with the proc amp predominantly, and I also use it to pull the saturation (chroma) down, though the 3-way colour corrector can do this perfectly well. For our shots of Gitta, I used this setup but found that it wasn’t enough.. the highlights weren’t strong enough, there wasn’t enough contrast in general. So for these sections of the doco I added a “Gradient colorize” filter — which can be quite radical. Suddenly your image goes completely black and white, with high contrast (as the brightest parts of your image are automatically boosted to full white, the darkest parts are assigned to full black). Obviously, we weren’t going for black and white, so the trick here is to pull down the “mix” bar, which effectively controls the strength of the entire filter. We pulled the mix down to around 6%, give or take. The result is that your highlights and shadows become stronger, you get more definition and contrast, but you’ve still got a colour image because the gradient colorize is turned down. It does tend to desaturate your image somewhat, but you can resaturate it using the chroma setting on your proc amp (which you put after the gradient colorize, since the filters are applied in hierarchical order).

And that’s basically our colour grade setup for the doco. We added a vignette to most shots using the 8-point garbage matte filter, quite a time consuming process, but as Paul said, it really draws the viewer’s attention into the centre of the frame and makes everything much stronger.

So yes, here’s the before shot:

And after.

On my computer screen, I don’t see a huge amount of difference. And everything coming out of fcp looks quite dark on my computer screen. But trust me. On the edit suite displays, it’s much stronger.
That’s all for now

Worst case scenario

The doco’s coming along well, I’m thinking. We’re putting in our overlay footage, and it to some extent works. It looks really interesting, and most shots are (somewhat indirectly) relevant to what Gitta is saying at that particular moment, without being too obvious. It’s an eclectic mix of visual material, and I’m thinking we’ll need to colour grade it in a reasonably uniform way to make it more cohesive. Should be alright.

But saying “Our doco’s coming along so well, we’re up to colour grading, I hope it looks nice”, etc, is ridiculously boring. I realise this. It would be more interesting, and also more constructive, to list my biggest fears about the way this documentary will turn out.

1. The core narrative progression will seem incomplete
We’re working on overlay footage too soon. We cut together a rough story progression extremely quickly, so that we could think about the projected imagery. But we never went back and seriously reworked our fundamental story. After showing the rough cut to various people, a pattern emerges: Our doco is split into two halves. The Holocaust is part I, Gitta’s philosophy on life is part II. What people want is for these parts to be integrated, or for the narrative to come full circle, return to the Holocaust and provide a direct resolution. We haven’t really attempted this.
“It already works. Why mess with it?”
It’s a mentality that’s hard to escape when you’ve worked for ages on 6 versions of the project, and the thought of completely reworking it is nauseating. But it’s true, our documentary ends somewhere completely unrelated to where we started. How can we make the story come full circle? Serious reworking. But as we layer on more projection footage, changing our narrative becomes less and less likely.

(One could argue that with the addition of strong visuals and music, flaws in the narrative will become less apparent. It’s less like reading a book.. maybe the sense of resolution will come through our audiovisual journey.. something like that. But then, why not have resolution in your narrative as well? That’s surely better than covering up your flaws with general “atmosphere”.)

2. We’ll run out of overlay footage
This may well happen, it depends on how much we want to hide Gitta’s face. It turns out that most of our projected material features the exact same shots, and we can’t just recycle the same imagery. I don’t mind the idea of letting Gitta’s face emerge more towards the end, as we’re getting to know her, as a sense of “Gitta” emerges from all the disjointed fragments of her home, her memories, the imagery, etc. We’ll figure this one out.. not a huge problem.

3. Our overlay footage won’t match our story
This is a big one. We’ve purposely avoided going for the throat and bombarding the viewer with blatant Holocaust imagery. I think it was a good move, I feel it gives us more credibility and originality. But what have we used to fill the void? We’ve got shots of Gitta’s house, the books she reads, her family photos, etc.. plus WWII warplanes projected over the walls, flowers covered in nondescript projections, time lapses of clouds moving overhead, intense sand animations, stabs of corroded film footage too damaged to depict anything in particular… yea. It’s different, and I like it. I’d wager we’ve got a lot more “eye candy” than most other TV2 docos, and at least that’s something. I want our film to stand out in some way. But in reality, a cool overlay shot will not save us if the film itself is flawed.

I don’t know anymore. Most likely, once we’ve gotten our overlays down and our soundtrack’s in place, our film will exist as some kind of interesting journey. Potentially flawed, but interesting. And this is documentary we’re talking about. The story doesn’t need to wrap up as neatly as it does in fiction.. because like it or not, this is what we filmed.. this is what happened.

Panic Room

The edit suite is quickly becoming a place of unpleasant revelations.. moments where you realise that your film is hardly as stunning as you think it is, that your footage is uglier than you think it is, that your participant’s story is less personal and more generic than you thought it was. It’s the crucial stage of our documentary production, and we’re learning that the narrative, the core of our documentary, is generic, impersonal, and needs restructuring. And also, we need to shoot overlay footage in Gitta’s home next monday, and it has to be vaguely cohesive once we’ve put it together, and match the feel of the piece in its entirety. There’s very little time to construct our own projected material, so the majority will be taken from youtube. This means we can’t vignette it, we can’t desaturate it or enhance it, and the quality will be highly variable, depending on compression codecs. Not a massive problem, but not ideal. Now while I was in an IM2 tute, the rest of my group brainstormed up some great visual material, and corresponding urls to the footage on youtube. Some of it’s awesome, and my philosophy at the moment is “the more the merrier”. A problem in post is the worst kind of problem because there’s nothing after post. After post there’s a black hole, a vacuum, the end of the universe as we know it. So we should prepare for this coming chaos and arm ourselves with as much visual material as possible.

So yes. We’ve got quite a lot of potential footage, and I may shoot a bit more on the weekend, if I get time. The more pressing question is: where do we project all this? Yes, we’ll do it in Gitta’s house, obviously, but after visiting, I’m thinking it may be difficult to find good projection-points. The walls are flat and white, but there are lots of large paintings on the walls that may make our projection shots look too convoluted. We need a blank space, a tabula rasa, somewhere in the house that i’m yet to find. We’ve also got to establish that we’re projecting within Gitta’s home. We’re going to start with naturalistic shots of Gitta’s house, various bits and pieces, fairly dark and atmospheric, but perceptually ‘normal’. Then we start fading in projected material within this space, and establish that these eclectic, dream-like images permeate the space of Gitta’s home, her everyday life. (This is sounding very IM2.. bachelard-esque.. no?).

But anyway. The main problem is that we’ve got all this material, and nowhere to put it. Perhaps we’ll have to resort to making a space on the wall, but I’d rather not, since we don’t want to be disruptive. Even if we did this, we’re not going to project all the images on the one surface, I don’t think. That would become boring, irrelevant, I don’t know. Certain things can be projected on the ceiling, in corners of the house, across the hallway perhaps. But we need to convey three-dimensional space, play with depth and visual texture… hopefully it all comes together.

I’m thinking that we should get static cutaways of Gitta’s house first, and then move on to the projected material. Otherwise we’ll be obsessed with getting the projections right and neglect the more ‘normal’ visual material. Just making a mental note of that, so that when we arrive on monday there’s some kind of concrete plan to follow.
That’s all for now.

Primordial Soup

I’ll tell you what I like about my group this semester.
We have disagreements and openly express our thoughts about different ideas. it doesn’t sound like much, but it’s crucial. After having so many meetings, shooting with each other on multiple occasions, and developing this doco idea collaboratively from the start, we’re comfortable enough to have a disagreement, an argument, without hurting anyone’s feelings. There’s no indulging in ideas out of politeness, we’re all openly concerned with what’s best for our documentary, for the final product. I think this is the way it has to be. Everyone’s free to voice any idea, everyone’s free to voice problems with said idea. Pretty much, anyway. There’s always some degree of inhibition, some treading on eggshells, some damaging of egos, in any group, you’ll find that. It’s never eradicated, but I think in this group, it’s definitely minimised. I feel comfortable in our group, and I’m fairly sure all of us are comfortable. I really hope so.
I don’t talk about collaboration enough, considering that film production is inevitably a collaborative process, and that we’re all social beings with ideas and emotions. Making a film is an organic process, it arises out of primordial soup.. the congealed accumulations of minds working together.. you can’t quite predict what’s going to happen, because you alone can’t determine how the film turns out. The film is determined by the relations between people.. it’s always complex, always interesting. I’m extremely curious to see how all the films I’ve worked on turned out at the end of the year… so many hours devoted by so many people, and it all becomes a few minutes of final product. I still can’t get over that.