• Cinedeck – super expensive and bulky
• Gemini 4:4:4 – expensive, inconvenient workflow
• Capturecards like the Ultrastudio 3D – cheap, not mobile, need to attach to computer and storage
• KiPro mini, Atomos Samurai, etc. – Only 4:2:2, not uncompressed

I decided to go the cheap route with this setup: Ultrastudio 3D hooked up to a thunderbolt raid and a macbook pro. In addition to this I also bought a Atomos Samurai mobile recorder. I chose this recorder, cause it’s cheaper and I don’t need the extra features of the PIX or the KiPro mini. The main reason I bought an additional on-camera recorder was for portability. Now I have a “studio-setup” with my Ultrastudio for capturing uncompressed 4:4:4, and a mobile setup capturing 4:2:2 in prores for documentary work etc.

Now where is the difference in quality? How does the 4:2:2 prores capture compare to the 4:4:4 uncompressed signal? Here is my quick and dirty test:

4:4:4 ungraded vs. 4:2:2 ungraded (hover over image to switch)

4:4:4 graded vs. 4:2:2 graded (hover over image to switch)

Here is a downloadlink to the fullres (1920×1080) images: 422vs444

Conclusion: To me, at first sight, they are almost the same. The 444 is a bit sharper and has a little bit more fine contrast but that wouldn’t justify the extra effort and expenses for shooting in 444. So far for primary color correction. When it comes to pulling a key for secondary color correction or for compositing work, there IS a big difference!

See how the keys/mattes differ between 4:4:4 and 4:2:2 (hover over the image to see the difference):

Difference when pulling a key in secondary CC between 444 and 422

This means to me, whenever it’s practical to use the 4:4:4 workflow, I’ll go for it. My colograding will definitely benefit from it. For greenscreen work there is no question whether using 4:2:2 sampling or not. It’s a simple no-go for me.
What do you think? Comments are appreciated!

Convergent Design announced „the perfect recorder“ for the S-Log/444 enabled F3 at NAB2011. The Gemini 444 is a hybrid monitor/recorder capable of recording dual-link 4:4:4 in uncompressed onto SSDs. With a price of $8.000 excluding SSDs and a formfactor not much bigger than a smallHD DP6 monitor a very attractive option.

I had several jobs from August to November on my F3 and always used 4:2:2 recording with a KiPro mini while waiting for the Gemini 444 to become available. Only thing I was concerned about was that the Gemini only records in uncompressed which means an enormous amount of data. (about an hour of footage = 512GB) So I knew I would have to bring a DIT-Station with plenty of datastorage to every set to backup and format the Gemini SSDs.

Now, three months later, with the advent of the Ultrastudio 3D from Blackmagic and Thunderbolt RAIDs I don’t see any point in buying a Gemini444 anymore. Here is my current workflow:

Click for larger view.

I am capturing the 4:4:4 output of the F3 into the Ultrastudio 3D connected via Thunderbolt to a 17″ MacBook Pro. The Blackmagic software running on the MacBook Pro is capturing in 10-bit uncompressed RGB and saving the files on a Promise Thunderbolt RAID5. Because the Thunderbolt RAID is so fast I can run an additional task while capturing. Adobe Media Encoder is watching the incoming folder on the RAID and is encoding the uncompressed files to ProRes4444. The finished ProRes files are saved on a 3,5″ SATA HDD docked in an eSATA dock. The dock is connected to the MacBook Pro via an ExpressCard eSATA adapter from Sonnet (that’s why it has to be a 17″ MBP for now until Thunderbolt to eSATA adapters are available).

With this workflow I end up with the original uncompressed file stored on the big RAID and a smaller ProRes (proxy?) file on the 3,5″ HDD. After shoot the HDD is sent to postproduction or given to the client. The uncompressed files are kept for final mastering, CGI and colorgrading.

See how fast the Promise Thunderbolt RAID is.

Compared to the Gemini444 workflow there is one big disadvantage: it’s bulky, big, heavy, and you need external power supply – no batteries. But remember: the Gemini 444 records onto SSDs with a max capacity of 512GB. The Gemini can span over two SSDs which gives you a max capacity of 1024GB and a recording duration of approx. two hours. When you shoot more than two hours of footage on set you have to buy additional SSDs or download the footage into a DIT-station which is practically the same thing as my workflow – a computer with a RAID.

With my described workflow you have several advantages over the Gemini444:

• you record uncompressed and ProRes (or any other codec you choose) simultaneously,
• you get plenty more of diskspace secured in a RAID interface,
• you skip the step of downloading the footage for postproduction,
• you can perform testgrades on set in AE or in Resolve,
• you basically have a very modular system and a fully functional computer,
• in addition to SDI you can record component, HDMI, XLR Audio, etc.,
• and many many more…

So what about the price differences? The Gemini without drives is about $8.000. Fully loaded with two 512GB SSDs it is about $11.000. Let’s see what my setup costs in total:

• Ultrastudio – $995
• SDI to HDMI Converter – $295
• Small HD DP6-SLR – $899
• 17″ MacBook Pro – $2499
• Promise Pegasus R4 4TB – $1149
• eSATA dock – approx. $50
• 1TB 3,5″ SATA drive – approx. $100

So everything you need to basically get the same as you get from the Gemini adds up to approx. $6.000. (I included the SDI to HDMI converter and the SmallHD monitor just because the Gemini can be used as a monitor too. If you don’t need that the sum drops to below $5.000)

Don’t get me wrong: I am sure the Gemini is a wonderful device and definitely has its uses, but for $11.000 it is not cost effective to my eyes.

Here are some pictures of my test-setup:

UPDATE: I made a quick test to compare the 444 uncompressed signal to a 422 prores recording. Here are my findings: http://blog.mxr.at/422-vs-444/

We wanted to shoot a high quality campaign-shoot for Fall/Winter 2011 and I just got my Sony F3 with the S-Log option so I gave it a spin. We took our whole setup including a modified Kessler Shuttlepod (one length cut into half to make it flight-ready), and traveled to Milan. My friend managed to get some professional models for the shoot, which raised the quality significantly I would say.

Shooting with the F3 handheld and on the Shuttlepod was great fun! Grading the S-Log footage is so nice. I’ve been used to DSLR footage in color grading for about two years now and having all these opportunities with log-Footage and 4:2:2 10-bit is simply amazing.

My only concerns were the lack of the S-Log option when shooting in 720p50 (the shot with the slomo doves), and the lack of a viewing LUT when recording onto the KiPro Mini. Everything is so dull and grey until you move the footage into color correction. I hope that my Gemini 444 will arrive soon as this recorder is going to have a monitoring LUT option. In addition to uncompressed 444 recording

Here are some pictures of our shoot in Milan.

Me with the F3 and our Shuttlepod.

The Zacuto EVF makes pulling focus a lot easier.

We wanted to give the production a nice cinematic look so we added our Kessler Shuttle Pod system and the Miller Projib to the gear we used. I was impressed how well things worked out. Especially the combination of the F3, S-Log and the Kessler Shuttle Pod with outrigger feet enabled us to shoot incredible slider-shots in the woods. When talking about latitude and dynamic range – I am satisfied with my S-Log pictures and I don’t need no HDRx.

Here is the director and my first AC working on the camera mounted on our shuttle pod.

As soon I have something ready I’ll post some footage.

I used the Sony PL-Mount kit lenses 35mm and 85mm and the Zacuto EVF. (If you look closely you can see Sandra carry the other lens)

Here is a picture Sandra took of me filming her:

I would like to thank Tools At Work, my local Apple/AJA/Blackmagic/etc. retailer for providing the KiPro Mini.

Next thing to check is RGB444. Now when I figure out why the SDI in ports on my Decklink HD Extreme 3 is not working,…

HDMI in and out for monitoring the 5D stream. As soon as you hit record on the 5D you lose the signal in the Hyperdeck.

So I recorded a color chart via HDMI in uncompressed first and then switched over to 5D internal recording method in h264. After that I imported both clips into after effects and investigated the blue channel. As expected the uncompressed signal was much clearer than the h264 stream.

Blue channel only. You can see the uncompressed signal is much clearer.

Here is another screenshot after a little curve correction was applied. All in all the uncompressed signal holds color correction much better than theh264 clip. The uncompressed clip has been upscaled by 118% to fit the original h264.

h264 vs uncompressed after curves correction

So my conclusion for this test is that there is a clear advantage when shooting in uncompressed even with the 5D. But to justify its drawbacks (need to upscale the footage, enormous amounts of data – 128GB hold 16mins!, handling, etc.) the only scenario where I will shoot this way with the 5D will be in greenscreen shoots. This is where you need this kind of color detail the most.

This is how I will shoot greenscreen with my 5D in the future.

After playing around with it about a day now I am quite surprised how much fun it is to work with. Here is a snapshot I took of our lovely doggy, colorgraded in Aperture:

And here is a video I took with the Canon 5D Mk2 and the Lensbaby Composer of my sister playing piano, colorgraded in Apple Color:

Music is by “We are wolves“.

With my Canon HDSLRs (Canon 5D MK II and 7D) I am shooting in the standard picture profile with the contrast setting turned completely down which gives the image a little more dynamic range. To lift shadow areas and crush bright areas even further I use the free Shadow/Highlight Plugin from Lyric Image Control. You can download it here
Since I have a quite flat image right now I can reapply contrast. 3-Way Color Corrector works very well for this but I am using an old preset from Magic Bullet Editors (which isn’t available anymore) called “LS Max Contrast”. Magic Bullet Editors has been updated to Magic Bullet Looks which also has a preset called “Max Contrast” but in my opinion they aren’t the same.
Last thing to do is to color correct skin tones. With the process of killing contrast and reapplying it, in frequent cases skin tones come out bad. I fire up my vectorscope which has the wonderful feature of a reference line for skin tones built in. (See the white arrow in the screenshot here)

With the 3-Way Color Corrector limited to just the skin tones I can push the skincolor towards the skintone line in the vectorscope. In the example above I had to push a little bit towards yellow. The result at first sight might look a little bit yellowish, but thats just because you have been used to looking at reddish skintones. Hold your hand up to the image and you’ll see that the skintones are correct.
The skintone line in Vectorscope is your friend!

One quick tip when color correcting skintones: If you want to limit the vectorscope display to a specific part of the image, just use the crop tool in the motion tab of the clip to crop the image to just the area where skintone is visible. That way you can work specific on your skintones and when you’re done just remove the crop.