Alexa post - Log-C, LUTs, best practices

I've got three calls since morning from editors in prominent post houses - one with offices in LA, Vancouver, New York... yeah, yeah we know... and another with offices in Bandra, Andheri, Chennai, London… 

Complaining that Alexa footage looks 'milky and low contrast'.

I had written earlier about struggles in digital movie-making

http://neilsadwelkar.blogspot.com/2011/05/struggling-with-digital-movie-making.html

And on Arri Alexa Avid workflow

http://neilsadwelkar.blogspot.com/2010/12/arri-alexa-avid-workflow.html

Some details on the Alexa camera itself here…

http://neilsadwelkar.blogspot.com/2010/05/arri-alexa-digital-movie-camera.html

Here are some simple facts…

The Arri Alexa camera can shoot 'linear' (technically called Rec.709) which looks 'normal'. Or as LogC which looks 'milky and low contrast'.

LogC gives you more latitude. Which, depending on the calibre of the cinematographer, means, he can mess up the exposure and you fix it in post, or he can shoot amazingly and you can change the 'look' and better it in post.

On the set you can shoot LogC (milky) and yet monitor Rec709 (normal). So, many people get fooled into believing what's on the monitor is being recorded.

LogC need to be converted before you can see them normally.

Avid has no method of doing this logC-linear ('milky' to 'normal') conversion.

In FCP you can get Nick Shaw's plug-in for $ 30 (Rs 1500) and apply it to all clips.

But it is 8-bit and not for production.

Alternately you can download a LUT for the Log-C to Rec709 conversion from Arri's web site. Google 'Arri LUT generator' and you'll find it. And also Arri's excellent white paper on how it all works.

You can also use Resolve, or the free Resolve Lite to do this conversion. Download it for free from Blackmagic's web site, struggle with it for a few hours, and then get an expert to do it for you.

Best of all, if you're shooting for TV, and you really don't want to mess with this Log-C business, just shoot linear or Rec709.

About the data coming out of the camera (can't call it 'footage' any more)

Alexa shoots HD 1920x1080 on memory cards - called SxS cards.

or it can shoot '3k' to a Codex digital recorder.

or it can record out to as video to an HDCamSR VTR on to HDCamSR tape.

On memory card it shoots Quicktime movies in the Apple ProRes4444 format.

On the Codex it shoots ArriRAW files that have to be converted before you can edit them.

Copying all this data to WD, or Seagate or LaCie disks is a risk. A drive fails, you need to reshoot. Period. 

There are two kinds of people - those who have had to reshoot because of a hard disk failure, and those who have yet to reshoot because of a hard disk failure. Carry a piece of wood to knock on from time to time if you are the latter.

Or, hire someone to do data management for you. Whatever you decide, don't let the post house manage your data. And never leave your data on someone else's hard disk. It is vastly easier to browse, copy and use anyone's data in a post house's shared hard disk - than it used to be in the 'film and video tape' days.

For data storage on movies I'm currently working on - Canon7D, Red, or Alexa, TV or cinema or commercials - I'm currently using RAIDs from Sonnet, G-Tech, and Maxx. All are good. And safe. I also advice backing up to LTO tape for long term storage.

On backup for all this media, I've written about this a while ago, but the basics remain the same…

http://neilsadwelkar.blogspot.com/2009/12/backup-for-tapeless-media.html

In terms of disk space…

Shooting HD as ProRes4444 uses up about 140 GB per hour of shoot, 

ArriRAW does it at 10 GB per minute, or 600 GB per hour.

ArriRAW has more resolution and latitude than ProRes4444, but at a cost.

There's only one Codex recorder that can shoot ArriRAW, in India, so far.

If all this sounds complicated, it really is just a bit. If you'd like peace of mind, do all that I've recommended and you're safe. 

CAS Day 2

Day 2 at the CAS show was one of seeing the same shots over and over again and watching for differences between one camera system and another. It mainly was focused around film vs. each of the other digital cameras.

If you don't know what CAS is, or haven't about CAS in Mumbai, head over to my earlier post on CAS. For a report on CAS day 1, head over here.

My broad conclusions from these tests may or may not match those of a cinematographer, because I ain't one. I saw each camera's output from the point of view of, can this tell a story. Whether it (anything digital) looks as good as film is something I stopped looking for a while ago. Nothing looks like film. But some digital stuff looks rather good.

So, some random observations...

There was a handout distributed at the start. It listed in a table the basic characteristics of all 8 cameras that were tested. Some of the stuff there could be contested. For instance, F23 and F35 are shown as uncompressed. Neither can record uncompressed to tape.

The Arri D-21 is shown as edit ready. But if you shoot ArriRAW then all that stuff needs to be converted to Quicktime before you can edit it. Just like with Red. But if you shoot D-21 to tape, then it is edit ready, but then, its not uncompressed.

These tests were shot in Jan 2009. And post completed in April 2009. Some changes have happened in digital cameras since then. Red one has been superseded by the Red M-X, and the Epic is near completion. Arri introduced the Alexa. And, of course, there are HD DSLRs.

On the test methodology and workflow...

Manufacturers were asked to do the conversion of their digital material to 10bit log DPX. And all work was done in this medium. In a real world situation if one is shooting digital, one may wish to work in the camera's own format. Or another linear format. 10bit log DPX is the preferred medium for film scans, not necessarily suited to digital.

Also, in these tests, the film shots were graded first. Then the 10bit log DPX converts from all digital cameras were graded to match the film as closely as possible. Only simple colour correction was used and no windows, or no sharpening or noise removal was performed. So in that sense it was a clean test.

One thing someone from the ASC mentioned was "Editorial was ignored". I guess that meant that the shots were simply plonked on a timeline with no relation with one another. In a real world situation, one always builds a scene as a series of shots. And maybe in this respect, digital is much more uniform than film. Meaning shot to shot variations are far less in digital than in negative film. meaning. Film absolutely has to be graded in some manner. The editorial factor has to be be part of the testing, As has sound.

Another observation I had was flicker was common across all formats. I mention this because I've lost count of the number of hours I've been debating this with commercial film-makers when the camera pans across a bright surface and it strobes and flickers. They blame the telecine, the colourist, the DigiBeta VTR, the FCP, even the editor. The CAS showed that ALL cameras flicker during pans.

When showing the tests, there were some observations expressed on the voice-over. I thought them to be leading, and it highlighted some defects that a cursory observation would have overlooked. After all no one watches one single shot over and over again, unless they are editors.

Archival in the digital age

It was narrated how, during the CAS shoot, one entire shot shot on the Viper camera erased because of a wrong interpretation of an alert message. This was used to highlight how in the digital age, its very easy to lose work. Whereas with film, it can't be erased.

I agree. Digital data is far more 'fragile' than film, when it comes to clumsiness and callousness. But out here, producers often skimp on hard drive costs, so they buy the cheapest one they can find. So archiving of digital data while production is on-going is a neglected area.

But, here in India its just as easy to lose film. We have lost negatives of films shot in 2003 to bad storage. We have this serious problem of heat and dust. And spotty electric supply. So, outside of Mumbai, even if you store your film in an air-conditioned chamber, where will you get 24x7 electricity? 

On the other hand, huge amounts of data are securely stored for extended periods in India. The Indian Railway reservation system is the world's largest database. The Census and UUID program, the largest database of human beings. Our IT boys store data for the world. Hollywood may not have the answers to data storage, but the Indian computer industry might know. We probably should ask.

On a side note, for most of the trash we produce on television and even in movies, do you really want to store that? Will people really watch any of it? Wouldn't you wish you could erase all that film and reuse it for something better?

Just go over the list of flops of this year and you'll see what I mean.

But yes, our classics, they need better storage, and digital is probably not it.

Conclusion

The overall feeling I got was that the differences between film and digital cameras of Jan 2009 was marginal. Cinematographers will debate on the finer points, but the truth was out there on the screen for all to see.

And, the reality of this test between film, Red, Viper, Genesis, F23, F35, D-21, 3700 for us in India could be summarised as film vs Red. We simply don't have any of the other cameras in large enough numbers (like exceeding 1-3 all India) to matter for film making.

On day 1 we saw film vs Arri Alexa. And day two was film vs Red. Then there are HDDSLRs which were not tested. So, in India, we need to make our choices between these. Film, Red, Alexa, HDDSLRs. For TV you can add EX1/EX3 or F950 type of cameras as well.

In the end, Kodak showed a film highlighting their new Vision 3 stock. it looked impressive. But the film also highlighted the role of cohesively edited material, a story and good sound to the overall film experience.

For the future, I wish the Cinematographers Combine, and some camera rental company, and a well equipped post house in Mumbai do a film vs digital test based on cameras we have here. And make it a fuller exercise with a story, direction, acting, sound, and editing. That's really real world.

I'm available. 

Mumbai CAS Day 1

The Cinematographer's combine CAS event got off this morning at a packed Imax theatre at Wadala. For a change the participants got there before time and the hall was packed with time to spare. I take this as a sign that digital is ready to be received. Or, that there was very little traffic in Mumbai this morning.

The first session was on the new Arri Alexa digital movie camera. I wrote about this camera on this blog a few weeks age and you can find it by looking through the links in the margin to the right. Or here.

The day began with a talk by Henning Radlein of Arri. I've been hearing Henning for over 5 years now. At various trade shows and demos, even a guided tour around Arri that he kindly gave me some years ago. And he delivers, with the manner of one of my professors of Classical Mechanics at Physics post-grad University class. Simple, concise and technically perfect.

There were some features that Henning spoke about that are not there in their brochure or web site. There was also a hands on demo of the Alexa, played out live on the big screen with a live camera. A rather good way of showing a device to a large audience.

Later we saw tests that Arri shot on an Alexa. These were shown as Digital, from the (presumably) Christie 2k projector. At 2k, or maybe 1080p24, I forgot to ask.

To my untrained eyes, the tests looked awesome. And I didn't find anyone who said anything derogatory about them, and most seemed suitably impressed. Until we saw the Indian tests later in the day.

The Indian tests were shot by two leading Cinematographers - Sudeep Chaterjee and Amit Roy. They had both shot some typical film scenes with props, actors and the like. And they both shot to film and Alexa and showed both one after another, as a film print.

I'm no cinematographer so a lot of the figures about shutter angle and aperture were just numbers to me. So I just saw it as one picture compared to another. From the general murmurings around the house, it seemed there were many who were doing the same.

One set of tests had a freeze frame in every shot with animated supers showing what seemed to be some kind of exposure level. I found that a trifle disturbing. I wonder, even among those who understood what those numbers meant, how many could retain that info. I think a set of stills on the Cinematographers Combine website would have been a better idea.

The tests were shown as a film print. On a film projector that was weaving and vibrating. So that was a bit jarring. So it was decided to see them as a digital print or DCP. One astute person in the audience asked for the Arri tests to be also shown right after the Indian tests.

For me that was an interesting watch. Seeing tests shot here, vs those shot by Arri both on a clean bright digital projector. What a vast difference in the sheer quality of images. Amazing. I don't mean to slam anyone here, but I think for quite a few persons in the audience the truth was before their eyes.

There was also a Q & A with some of those associated with the tests. Unfortunately, out of the two cinematographers who shot the local tests, only one could attend and answer tests on the shooting. And, as far as post was concerned, the colorist was absent too. And Ken, who filled in, did a reasonable job of providing info.

I found the workflow for the Alexa tests locally done to be... well, 'interesting'. At Reliance MediaWorks.

They scanned the film at 4k. On a Spirit 4k. Graded at 4k on a Baselight. and recorded back to film at 4k too.

The Alexa shot to SxS cards as ProRes 4444 as well as HDCamSR as 444HQ. In the logC colour space. But Reliance chose to capture the tape. This was captured to Clipster and exported out as uprezzed 4k DPX, for grading in Baselight.

So, the film images were presumably 4096x1744 pixels during scanning, grading and recording.

The Alexa images were shot at 1920x1080, cropped to 1920x817, Scaled up to 4096x1744, graded at that size, and recorded to film.

So, if my presumption is correct, we were shown native 4096x1744 film vs 1920x817 size digital images side by side but on film. All those who debate on 4k vs 2k for film DI need to take careful note of this in relation to what they saw on the screen this evening.

To be sure, I asked Henning whether the camera provided to CC for tests was capable of shooting ProRes. He said Yes. So, the decision to use the tape instead was taken by Reliance. Maybe because they don't have a Mac at Reliance with the latest FCS 3 which is needed to open ProRes 4444. Or maybe they feel tape is better.

I also checked with Henning if the Arri tests that we saw earlier were shot to HDCamSR or ProRes. He told me which were HDCamSR and which ProRes. He also explained me their workflow, which was radically simpler and arguably more edit friendly that the one Reliance adopted.

For me, the highlight of today was how different our shoot looked compared to the one done by Arri. Exactly why, is something I'll ponder about.

Later, there was a presentation from Sony on the SRW 9000 camcorder. Actually, calling that somehow puts it in the same league as an EX3. but not really. This is a digital movie camera with an HDCam SR recorder built in. The SRW 9000 is based the F23. And there will be a SRW 9000PL based on the F35.

The SRW9000 will have a memory card recording option in some months. This is a flash memory card that will be 1 Tb and record uncompressed. This memory card option will also work with the F23 and the F35. There is also a 4k upgrade path for the camera section. Which makes sense as this special 1 Tb cartridge will be capable of 5 Gbps.

Sony and Arri are adopting similar approaches, that of recording to a know medium and format. Easy to store, edit and manipulate. Both are good cameras and I wish there was a Alexa vs SRW 9000 comparison instead of film.

For us in India, unless post houses use the Direct to Edit Apple ProRes function of the Alexa, both will become HDCam SR cameras. So rentals will be similar. It will boil down to which camera cinematographers prefer.

Eventually the 1 Tb SR2.0 card will be introduced. And some may demand the raw recording capability of the Arri Alexa or uncompressed for the SRW 9000 over ProRes or HDCam SR. But both these approaches take up huge amounts of disk space so they might be non-starters in our context.
 

The rushes of an average Hindi movie, uncompressed (SR 2.0) or ArriRAW will take up 12-15 Terabytes disk space. I know exactly how hard it is to get someone to invest Rs 35,000 that it costs to get one secure hard disk to store just one ad film's a Red rushes, so asking for Rs 12-15 lakhs ofr a hard disk array for rushes... forget about it.

There was a very basic test of some shots with the SRW 9000 shot a couple of days ago. These were shot as S-log. The camera played the tests directly with and with a 'linearizing' LUT applied. I felt the test was playing at double speed. Meaning, they were shot 1080p25 or 1080p24 and the camera system was set up for 1080p50 or 1080p60.

If it was, then that shows how careful we need to get with these HD formats. 

By the way, where does Red figure in all this? It does. And I plan on doing a compare, very soon.

I've written on the Arri Alexa, SRW-9000, and SR 2.0. Check it out. And please, if anyone is using this info for any commercial gain, consider the donate or subscribe button.

Arri Alexa Digital Movie Camera

At IBC last year, Arri announced a new camera. A successor to the D21 - a digital camera to shoot movies with. That was Sept. 2009.

 

Last week at NAB (a large international post-production exhibition and conference at Las Vegas) Arri showed the camera. Its now called Alexa. Arri had a working prototype at their booth. Something you could touch and feel and shoot with and watch the output of, on a HD monitor.

The most important thing Arri have done with the Alexa is to sort out the production and post-production workflow in a manner that's simple - probably simpler than the Red. Whether this helps them sell well and better then Red, only time will tell. The pricing is not a bit higher than the Red, though.

In it's base configuration the Alexa shoots at a resolution of 3 k with a frame size of 3072x1728. This is downsized to HD at 1920x1080 and recorded to SxS memory cards as Apple ProRes422HQ or Apple ProRes4444. ProRes422HQ has a data rate of 145 Mbps while ProRes4444 has a data rate of 330 Mbps.

To put this into perspective, Panasonic D-5 tape stores video at 220 Mbps and Sony HDCamSR does 440 Mbps. So, in terms of data rate, and hence arguably, quality, the Alexa at ProRes4444 is somewhere between D-5 and HDCamSR.

This HD output is also available as a HD video signal which can be recorded to HD video tape like HDCam or HDCamSR. In fact, this is a safe way or working. ProRes4444 on SxS cards and the same HD video to HDCamSR tape. The tape becomes a backup.

For even higher quality and latitude, the Alexa also can record Arri RAW to a certified Arri T-link recorder. Like the Codex, S.two, or KG models of recorders. These recorders cost as much or more than the camera.

When the Alexa makes its way into India, the scenario might look like this...

The camera itself (pro set with DTE option) costs about Euros 52,500 (about Rs 40 lakhs incl. Customs Duty). And the recorder that can record Arri Raw costs another US$ 60,000 (about Rs 36 lakhs incl. Customs Duty).

The Arri Raw recorder records to hard disk about 15 Gb per min. So, if you are shooting a feature film of 120 mins at a shooting ratio of about 5:1, you need about 9000 Gb or 9 Tb space to store 10 hrs of Arri Raw rushes. At a shooting ratio of 10:1 which is what people do, when they shoot digital, you're looking at 18 Tb. To be safe you need to back this up in two sources - disk and LTO.

Now, going by what non-film cameras like the Red rent out for in India, my guess is that the Alexa will also rent at a premium. Meaning you will need to shell out two times or more per day for an Alexa what you would, for an Arri 435 film camera. This differential alone will make potential film-makers think twice about shooting digital.

Add to that the FUD (fear, uncertainty, doubt) factor around digital, it will take a while for someone to have the guts to do an entire mainstream film on a digital camera like the Alexa. As if that isn't enough, the sheer cost of providing safe secure storage for digital Arri Raw data, will make most producers balk at the idea.

At a conservative estimate, 8 Tb of safe secure RAID storage will cost at least Rs 1,20,000 for an 8 Tb RAID. That will add up to Rs. 2,40,000 for 16 Tb. This cost will need to be borne by the producer as well. Of course after the film is done, one can delete the unused rushes and then reuse the storage for forthcoming films. But the initial acquisition of a high quality disk storage will have to be made.

Many film-makers may settle for Apple ProRes4444 which takes up just 130 Gb per hour, or 1.3 Tb for 10 hours going up to 2.6 Tb for 20 hrs. And, for many purposes, ProRes4444 provides nearly all the bandwidth and sheer picture quality that Arri Raw would have given you, but at a fraction of the disk space and bother of file conversion.

At NAB, I was told, about 300 Alexa cameras were sold worldwide. In all probability, one or more people from india have bought one, or are considering one. So, come June, when the Alexa ships, one will see an Alexa somewhere in India, most likely in Mumbai.

For someone looking at an alternative to film, and who doesn't want to mess with the workflow of Red because his editor or cinematographer has heard bad things about it. And needs a camera that responds just like a film camera - the Arri Alexa is a good choice. If you have the budget and inclination for shooting Raw, and a recorder is available, then go with shooting Raw.

The complete Arri Alexa specs ...

There's, of course, more to the Alexa than just post-production convenience. Its latitude, dynamic range, 35mm sized sensor and compatibility with existing 35mm film lenses. Features that might appeal to the cinematographer and make the camera closer to a film camera. There isn't (yet) an optical viewfinder, but they may add one later in the year.
 

But for ease of use and simplicity, stick with shooting ProRes4444. The files will open and can be edited directly in FCP. And even in Avid Media Composer ver. 5.
After editing, these files will open directly in Smoke on a Mac for any effects work. Even windows apps like 3DStudio Max can use these files for any CGI work. For grading, the yet-to-be-released Resolve on Mac will be able to grade ProRes4444 files.

So there you have a clean and simple workflow. For shooting digitally for features and shorts. Even ad films and music videos.