jcs

Many folks thought it looks excellent- did you watch the whole thing? It's typically not possible to get that much clean detail in low light with H.264 (usually turns to blocky artifacts and is hard to clean up, even with Neat Video. Note some of the clips are pretty clean due to NR, however it still has detail on edges so it looks pretty good for the conditions). Perhaps post a D800 wedding video so we see can compare?

Part of the learning process as a filmmaker is to try different cameras to see how well they work to and see how they cut together. There is no camera (yet!) perfect for all situations. Sometimes the camera you have on-hand is the best camera (or when the batteries died on your A-camera :) ). Here we tested 5D3 raw, processed with raw2cdng (very fast but no vertical stripe removal yet) and ACR (first shot, outdoors at golden hour), Sennheiser G3 wireless, Oscar Sound Tech OST-802 mic (great value at around $110 for the kit wired for the G3)), FS700 + SpeedBooster, and iPhone 5S for the last shot (FS700 batteries gone). Canon 24-105 F4L used on both cameras, same mic. Lighting (except first shot- only sun, no reflectors) was 2 6-element 'twisted bulb' high CRI fluorescent studio lights (sorry, don't have brand info anymore- they were low-cost) with diffuser material draped over lights. When the host (Elena) saw the iPhone 5S footage, she exclaimed "I like that better than the Sony!", lol. Getting good color out of the FS700 is possible, but takes work (other folks have stated the same thing- even with a 7Q it's not as good as Canon (or Apple!) color processing (from what I have seen online so far)). Here's 100% FS700+SB + Canon 24-105 F4L at F4, handheld with an HDV-Z96 LED light on camera. This was our first time shooting this kind of thing and the live audio ended up needing to be replaced with ADR (live quality was excellent with the Audio Technica 4029 stereo shotgun and Sony preamps, however the dialog had to be re-done). We used a Shure SM7B and a Focusrite Scarlet 2i2 for ADR (great preamps for the price- sold my RME Fireface 800* after getting this!), recorded directly into PPro at my desk in my home office (not a sound-treated room). I had not done this before and it was good practice (didn't try for perfect sync + some of the words were changed- kind of reminded me of old Clint Eastwood Spaghetti Westerns). * The Sound Devices USBPre2 has more accurate preamps, while the Scarlet 2i2 is a bit smoother and well suited for vocal recordings. The Fireface 800 is a great, very stable product, however I never liked the preamps (and while the routing complexity will blow your mind (in a good way), it's so complicated it will blow your mind (in not a good way when things don't work :) )). Here are the same clips on youtube: https://www.youtube.com/user/TheHotDishShow There was a time when youtube and vimeo offered similar quality- it's clear vimeo is now better (too bad as youtube has more social/virality features). We just shot another piece using 100% 5D3 RAW + Z96 light and G3 + OST-802. When we ran out of card space (2 64GB and 1 32GB Komputerbay), I pulled out the slower CF cards and continued with H.264. With good lighting I believe it will cut well with RAW, especially for internet delivery. I picked up this viewfinder and mini rail system for $130: http://www.amazon.com/AUTHENTIC-KAMERAR-VIEWFINDER-FINDER-NIKON/dp/B00EAXU6UM/ (says Nikon but works perfectly on the 5D3). I almost got the Zacuto for $300 but decided to try this first. I mounted a handle using the rails under the camera for balance: this setup works very well! Can't imagine the Zacuto or other brand would be any better in terms of optical quality and magnification ratio (2.5x)(Zacuto has anti-fog- not an issue so far in SoCal). Focusing wide shots is still an issue, even with ML peaking (have to punch in first). Really need an electronic viewfinder with full 1920x1080- I can easily focus when hooked up to an HDTV (no peaking even needed).

Canon (and the 5D3) has the nicest color processing in the price range, especially for skintones. In good/bright light, the 5D3 H.264 sharpens nicely and looks really good on a large HDTV at a normal viewing distance. It's very important to use sharp lenses and to nail focus. The 24-105 F4L is ok for closeups, but not so good for anything else (sharpness). The 16-35 F2.8L II, 24-70 F2.8L II, and 70-200 F2.8L II are very sharp and really help the 5D3 H.264 look its best. Nikon and Zeiss primes are also very sharp; the new Sigma 50mm should be spectacular. Canon L lenses with H.264 (IPB): looks good on HDTV: I also shoot with the Sony FS700 and Speedbooster. The FS700 is sharper and has greater dynamic range (at least 2 stops) along with smaller H.264 files. When white balance is set properly, color processing is decent but still can't match Canon, especially for skintones. The Odyssey 7Q (haven't used one yet) takes the FS700 to another level, though from footage posted so far the color processing is still not as good as 5D3 14-bit raw. The GH4 has resolution to spare but falls short on color processing and DR and has a tiny sensor (only S35 with a Speedbooster- an issue if needing full frame FOV). The A7S looks to have excellent color processing and DR though has a fair amount of rolling shutter. For run and gun, the 5D3 is better than the FS700 as it's lighter and easier to focus when using a quality viewfinder loupe. The FS700 can use autofocus lenses but they are slow and don't always focus on the desired object. 5D3 14-bit raw is spectacular, especially when processed with ACR. The tradeoff is card and disk space and processing time. For the price, this is the closest one can get to an ARRI Alexa/Amira- the best cameras in most people's eyes. (GH4 can't compete with this color; processed with Resolve- ACR wasn't used) http://www.dvxuser.com/V6/showthread.php?322289-Wedding-(RAW)&s=6bee780c6bd52fdfdd928bcee7cec245 (Low light raw, processed with ACR).

As other have suggested or implied, and has been done on many other sites- recruit moderators to help you (I was, and apparently still am, a moderator at Cinema5D, however I don't have time to be an active moderator). Perhaps from your selected helpful contributor list, ask those folks to also be moderators, and give them the authority to warn users and delete/edit posts. I believe that's all you'll need to clean up the forums.

If grading in a 32-bit float app as with most modern NLEs, the 8-bit integer source will be converted to 32-bit float- it won't really matter if you edit in a 4k sequence or 1080p sequence: nothing magical will happen regarding grading latitude if downsampling first other than perhaps higher performance.

We are all in agreement that we'll get 10-bits of luma information and 8-bits of chroma. Now, think about photons hitting the sensor. If the photons hit the 4-pixel group evenly, or with minor variance, we won't get much out of our 10-bit final result. Interestingly, noise will help get more variance. Will the noise act like dither or will the low-pass filter effect just smooth the result? Intuitively, I don't think there will be enough variance for the 4-pixel subgroup to have much of a visible effect in helping to reduce banding or in increasing grading latitude vs. 10-bits from the sensor. However, if you guys are having fun exploring through experimentation, carry on! :)

I don't want to rain on the parade here (I championed post-sharpening for the 5D3 when folks didn't believe it would work), however the math doesn't actually predict much of an improvement for this process. First, we're not going to get a 10-bit 444 file. We'll get (at best) an 8.67-bit 444 file. Regardless of credentials and title and who said what, it's not possible to get 10-bit 444 from 420 8-bit (unless perhaps the 420 is taken internally). Here again is what I wrote in the other thread (I'm a filmmaker, image-processing app developer, and game developer: I'm not a mathematician either :) ):

http://www.pcworld.com/article/2137477/samsung-lowballs-the-4k-competition-with-700-display.html Want 10-bit 1080p 24.5" OLED instead? http://www.flandersscientific.com/index/cm250.php ($6500) A 10-bit display is needed to see 10-bit material. For 8-bit displays, the image needs to be tone-mapped and dithered from 10-bit- this usually happens on final render. Not clear if apps like Resolve and AE do real-time tone-mapping+dithering to 8-bit (or at least when paused).

The A7S example 4K video looked, really, really good. Perhaps it's helped by the full frame look and nice bokeh. I saw some noise on the drum in the temple, however the skintones and overall image quality looked better to me than the GH4 examples posted so far. The A7S material looks a lot like 5D3 RAW, with much cleaner, higher resolution! Regarding 8-bit vs. 10-bit and real dynamic range: in my tests with 14-bit 5D3 raw vs. the 8-bit FS700, the FS700's ~14-stop sensor works better in high DR shots vs. the ~11-12 stop 5D3. A question is how does the internal 1080p look in the A7S vs. GH4 4k => 1080p in post. For folks who have Canon EF lenses, all that's needed is a high quality (Metabones etc.) EF to E-mount adapter to get full frame. A Speedbooster gets the GH4 to S35 (full frame not possible). The 720p 120fps S35 crop on the A7S could be full frame with a Speedbooster. A very interesting race between these two cameras: at the gate: GH4 4K resolution with internal recording Higher bitrates Higher resolution slomo 10-bit 422 external 4K A7S Color and image processing Low light performance Full frame DR especially with SLog2

Hi FilmMan- your eyes are workin' pretty well! :) Right, the suggestion was to put ARRI Alexa footage in the same clip as the GH4 and 5D3 RAW, etc., all in one ProRes 1080p file for download comparison. Additionally, shoot the same scenes with all the cameras (and ideally similar or same lenses). For the price, the GH4 is a rad little camera, however as djrockadoo points out, it's not in the same league as the Alexa, or even the FS700+7Q (which does excellent 4K 12-bit RAW, as well as in-recorder downsampled 4K to 1080p in ProRes; soon 4K/2K 12-bit 444 ProRes). Are those other cameras and gear worth the extra cost? For paid work, absolutely. Internet tests and cat videos? That depends on the owner :)

The 4K ProRes files are 10-bit 422 after transcoding from the original 8-bit 420. However, there's only 420 8-bits of color information in those 422 10-bit files.

Let's review the math: 10 bits is 2 more bits vs 8, 2^2= 4, so adding 4 8-bit values together gives us a 10-bit result. If we add up all the values, then divide by 4, thus averaging the result, we'll still see the extra information in the fraction. So if we average the values together in floating-point, we've achieved the 'effect'. This is effectively what an NLE will do when rescaling, so we don't need any special extra processing for an NLE that works in 32-bit float. 420 AVCHD (H.264) is YUV. If we scale 4K YUV 420 to 2K 444 YUV, only Y is full resolution, and only Y will get the benefit of the 4-value summing and additional fractional bit depth. Luminance is more important than chrominance, so that's not so bad. So best case, we have 10-bits of information in Y and 8-bits for U and V. This is (10+8+8)/3 = 8.67 bits per pixel. If the NLE does the scaling after converting YUV to RGB, it's still 8.67-bits of information per pixel at best (no new information is added during the transformation). This is why we won't see a significant improvement in color depth. Here's a challenge- create an example that shows the '10-bit' effect is significant (I agree it's there, but at 8.67 actual bits, it will be hard to see in most cases).

So for your test you can provide a 10-bit ProRes download ;)

Sorry Andrew, the 4K 420 => 2K 444 math of 8.67-bits/pixels (not 10-bits) doesn't support any significant extra color depth. Here's how you can prove it to yourself: grade something in 4K that starts to break up due to limited color depth. Downsample to 2K- see any improvement? 2K 444 8.67 bits is still very nice at this price point. Since the camera supports 10-bit output, perhaps a future firmware upgrade will support 10-bit H.264 (supported by the spec and used in Sony's XAVC). It might not happen for a while due to upline cameras such as the S35 Varicam, but should be possible.

Thanks for the test Andrew. You've convinced me on the next camera to get. The ARRI Alexa ;) Perhaps rent or borrow an Alexa and do a comparo such as this: Note how many folks could not figure out which camera was which.

PPro, AE (settable), and Resolve all convert to 32-bit float after decoding. Unless another tool does a better job of converting 420 to 444, there really isn't any advantage to transcode to 10-bit ProRes except for perhaps editing speed.

Sounds about right dhessel- what we'd expect from 8.67 bits and no sensor-provided increase in DR. HurtinMinorKey- is there an error in the math? Only green gets 2x unique samples in the Bayer array, red and blue are 1/2 resolution; Y is R*.3 + G*.6 + B*.1; only Y gets 4 samples to combine vs 1 for U and V in 420. If on-camera debayer is very good, the low res R and B will be better than 1/2 resolution, however at best we are still looking at 8.67-bit YUV after downscaling. After conversion to RGB, it will be somewhat less than 8.67 based on how good the debayer method performs. Thus not being able to see a difference matches what the math predicts.

The 4K to 1080 is (8+10+8)/3 = 8.67 bits; what you're seeing in practice makes sense. I too saw limited DR when playing with these clips. Masking for the sky and applying noise can help reduce the banding.

My comment was for PPro editing... slowdown was caused by adding just two RGB Curve effects. Changing PPro to 1/2 playback resolution got back to real-time. Resolve does 15fps with 2 nodes active (and modified curves). The PPro update I posted in another thread should help with 4K work. Comparing the 4K GH4 video to 5D3 RAW 14-bit and FS700 8-bit, the GH4 appears to show less compression than the FS700, however dynamic range appears a bit less than both the 5D3 (very close, but can bring up shadows well in post with 14-bits) and FS700 (guestimating by ~2 f-stops). Not conclusive, of course, just examining this initial GH4 footage (lots of clipped and/or quantized highlights (when brought down)). Someone mentioned IR and too much red- I see that too.

https://blogs.adobe.com/premierepro/2014/04/thenextpremiereproccrelease.html?sdid=KLDNF&PID=3740040 "Coming soon" - not clear when. Performance improvements should help with 4K. More elements coming from AE into PPro. Thus no need to go into AE for certain elements, however "When more finesse is required, simply jump into After Effects for greater control". That's like saying jump into a tar pit to provide finer control of your legs. AE can't even play blank video with audio in real-time ;). AE is a bit of a different editing paradigm, however it would be nice to see if PPro and AE could be merged into one app. For example, AE style sequence/comp editing being able to be placed in a PPro style sequence (similar to using Dynamic link, but never leaving PPro and getting 100% GPU acceleration for everything).

2010 Mac Pro- Bootcamp (I run OSX native on the MBP laptops, with VMWare Fusion for Win7 (utilities only, no major apps)). Running Mercury SSDs on the MacPro- wasn't a disk issue. Turns out adding just two RGB Curves effects slowed it down. Switching to 1/2 res for playback got it back to real-time. Guess the Quadro 5000 is due for a refresh (likely with a hacked consumer card from here: http://www.ebay.com/sch/macvidcards/m.html e.g. 780, Titan). Would be good to know if upgrading the GPU will help or if this is a limitation of PPro CS6 or CC (tested in CC). [Edit]: Resolve runs 15fps (resize set to Optimize for Playback, 13fps otherwise).

Can you upload original H.264? Looks like H.264 decodes faster on Windows at 4K with PPro vs. mov (slow QT decoder?). No advantage in this case to transcode...

The 4K footage looks great on a 30" 2560x1600 monitor (guess it's time to look for a 4K monitor). Running on Windows 7 (MacPro with Quadro 5000), the 4K mov's aren't real-time (QuickTime decoding, looks like). The MP4, which is much more CPU/GPU hungry, plays back in real-time (looks like native/GPU decoding). I'd guess on the Mac side the 4K mov's will play faster. Here's a quick 1080p grade on the 196x Porsche 911 using only RGB curves (including using my top-secret Panasonic to 5D custom curves preset ;) ): Original Graded Really dig the quality- looks like a shot from a nice stills camera.

Hey Andrew- looks pretty good, perhaps Vimeo "upgrade" the video to 1080p (for folks without Plus etc. accounts; I'm downloading the 1GB 4K MP4 now).

Sean- yeah, with advanced CPU/GPUs running on embedded devices, improved face/feature tracking will help focus even more. Infrared 3D cameras as with XBox connect could also be very useful for focus (a visible light stereo camera system with real-time image correlation (building a 3D depthmap measuring distance from sensor to Z-planes) could also be extremely useful for focus). Adding a giant distance display (that updates quickly) to ML would help with 5D3 focusing (with practice and knowing distance to target will be more accurate than peaking, etc., especially for wider shots). Hey John- I'm practicing to become a better filmmaker- your thoughts on my 5min short (in signature below)? (to keep marginally on topic ;), all shots focused looking at the 5D3 display (no loupe or external monitor)).