SONY a7siii + Arri Log C = $$$
HTML-код
- Опубликовано: 19 окт 2024
- Showing different methods for converting your Sony a7siii SLOG footage into more usable starting points culminating with Arri Log C techniques.
Link to Sidney Diongzon video:
• Sony A7S III Unboxing ...
Consider becoming a patron!
/ tonydae
Join us on Discord!
/ discord
Music credit:
Bensound.com
![Chino Pacas - Smith (ft. Junior H & Fuerza Regida) [Video Visualizador Oficial]](http://i.ytimg.com/vi/BVK6k56X30Y/mqdefault.jpg)
That was VERY helpful. I like the one you did for taking BMPCC4K to the Arri look as well.
Really like that video too. Finally got me to some grades I liked. I've watched it like four times haha
@@OneCheapEric I rewatch videos all the time. It helps me memorize the techniques. I also download the videos using Clip Grab. It saves the videos as HD to my computer.
Thank you! This is elegant
Thank you! This just made my 8-bit Sony footage look good!
Thanks Tony! Just got the A7c and am playing with SLOG3 - this is a fantastic workflow and I am happy with what I am seeing on my footage!
Good to hear!
Whow! That Alexa-Space transform rocks with my S1H footage shot in V-Log. Especially does this work flow bring back some highlights in a very pleasant way. Thanks so much for this huge hint!
You can do this trick with pretty much any footage. You can even try it with rec709 footage to take advantage of the shoulder provided in the Arri>rec709 LUT.
super video. there's nothing like actual screen recording of just doing stuff and doing it the right way.
Awesome video again ! Your channel deserves more subs !
Thank you, I really appreciate that!
Wow , very informative and to the point
Would a professional typically choose to grade from log to Rec. 709 themselves, or is it optimal to just use a Log-709 Lut and save time?
Disregarding color science, if someone wanted to maximize dynamic range, should they do a Raw to Log conversion into Alexa's Log-C, SLog3, or Panasonic's V-log? Those 3 Log curves seem to be highest in DR (Canon's CLog sucks). My camera's sensor is Panasonic Lumix, so perhaps V-log is designed specifically for that sensor.
@TN
Rec709 Manually V Color Managed or LUTs
---
As a professional I would say manually grading to rec709 is a waste of time. The rec709 conversion, either through a LUT or through Color Management, gives you a starting point and makes sure that the image is at least converted "correctly" so you can see what was in fact shot on your monitor correctly. Manually doing it leaves lots of room for human error that doesn't really need to be there at all. If you don't want to use LUTs, use Color Management instead.
Dynamic Range
---
Maximizing dynamic range is dependent on what you have available really. Some sensors and the imaging pipeline will simply not allow more DR than the log profiles provided in camera, so in that case, recording raw will not increase dynamic range in a useful way. If the sensor is only capable of providing 12bit video signals (and not 16 like Arri or higher end cine cameras), then recording in raw or even 12 bit might not yield the results you thought they might. Diminishing returns are more obvious when comparing less expensive mirrorless to more expensive cine cameras (compare say GH5 to Arri LF for example).
You will likely gain in compositing (background removal) and the ability to correct the image better with less destruction (shooting in the wrong white balance or ISO), but not so much in DR.
FX3 as an example, the difference between raw and log3 internal is higher ability to fix the shot in post if you messed up entirely and slight detail and color fidelity increase. If you shoot correctly, this is all nullified. Proresraw is not compatible with Resolve so its kind of not worth it at all in my opinion.
The only times I have seen much of a gain in DR from shooting in raw compared to an appropriate bit depth log profile is when you can use a composite to approximate clipped channels or the log profile is not capable of higher DR. The Blackmagic Design cameras have a recovery tick box in raw that can allow reconstruction of highlights that got blown in one channel by compositing from the other two. It doesn't always yield great results, but it is a thing you can do.
N-log has a DR limitation that truncates what the cameras can actually achieve, so you will probably see a DR boost with raw.
V-Log is excellent and transforms very well into rec709 with color management in ACES and RCM, and you can use the same LUT workflow used here.
Please feel free to reach out to me personally via my email address located in the ABOUT section if you have further questions.
@@TonyDae Thanks Tony. I'm new to most of these things, so this is rather educational. I already have a Lumix S5 and Atomos, so my workflow will be involving transcoding the Raw into Pro Res 444 using SCRATCH (to edit in Resolve). I have the option of Raw-Log conversion in Scratch. I am wondering if I should choose my manufacturer's Log curve V-log, or Alexa's Log-C in the conversion, to squeeze out the extra bit of fidelity I can get (I'm not sure which to choose based on maximum DR benefit). I also suspect maybe it is safer for me to just pick V-log since the log is designed for the Panasonic's sensor, although I do prefer the more film stock look of the Alexa log.
@@PunishedSeeker I don't think it'll make any difference provided you work in a color managed pipeline and the conversion done in scratch is done correctly. Both should be able to hold the information well enough in 12 bit 4444 and the rec709 conversion should look the same with little difference, provided it's mathematically done with color management. The only difference might be how they operate with control in a non color managed pipeline and I'll say logC is easier to manipulate in my experience. The best bet will be to shoot several test shots and convert to each then compare in a stress test.
Hi Tony, your channel is a gold mine, thank you so much. Is there a difference at the end of the tree if we use a cst arri > rec 709 instead of the arri lut ? If we use 2 color space transforms, where do you choose to map the luminance and the saturation, in the first one or the second ? Thanks a lot.
The transform will be different. Map Luma values in the second cst node. Color is sometimes dependent on how well the source color gamut maps into the working color gamut. If you're seeing gamut clipping in your working color space, you'll need to do at least a little saturation mapping in the first to prevent issues. If colors are going outside of rec 709 after the second CST, you'll need to do some mapping there.
You can also try running your source through ACES. The new ACES gamut compression makes using ACES a lot easier.
@@TonyDae Thank you so much for the advice. I will try it soon and experiment.
Feel free to email me if you have any trouble. Address in the about page.
Dude! OMG! You just made triljons of resolve tutorials completely obsolete! I was never a fan of LUTs, slapping one on footage does not make for a good learning experience. But transforming color this way is just plain genius! Every test I did so far was a breeze to grade from this point on! Awesome!
I have a color managed version that works through ACES and provides these same results regardless of camera and output color space coming soon!
If you want a different look, try changing the color space node outputs to rec709 or arri color space with cineon log and replace the arri lut with a rec709 lut found in the film looks folder that comes with Davinci Resolve.
Tony, thanks a lot for sharing your knowledges.
Thanks Tony. On the last node that you labelled 'LUT' and applied the Arri LUT (Alexa LogC to Rec709), I'm just wondering, instead of applying the LUT, can we just use color space transform again (Arri Alexa / Arri Log C to Rec.709 / Rec.709)? I imagine they're doing the same thing, but like you've mentioned before, color space transform would be less destructive? Thanks!
Yes you can, but the results will be different. The lut has a particular way of handling color and luma data that gives a nice initial look without having to do much. Of you use a cst, you will want to make adjustments to color and luma output for best results.
@@TonyDae Thanks Tony, that's good to know! Appreciate you sharing your knowledge
Learnt a lot here
Hey Tony,
Been doing grading for a while but only recently started looking at more scientific / standardized workflows. I sometimes round trip to Davinci for bigger projects, but for smaller ones I use FCPX. Was wondering what the equivelant workflow is for Final Cut with the color space transform, or if that doesn't exist in FCPX?
I haven't used final cut since maybe 2008 so unfortunately I can't speak to a color managed workflow in that program, however, If you do a node based workflow in Resolve you should be able to output a LUT that you can use to get the same transform in fcpx.
Good stuff Tony!
Awesome!
a lot of A7siii users seem to use Phantom Lut for Arri Alexa look. Have you used it? Wondering how the lut compares to your method of grading
I can't imagine any faux arri lut doing anything other than a transforming the sony color/gamut into arri color space and simulating their manufacturer rec709 gamma transform. Anything else would be a specific look transform by someone using arri footage since arri cameras are capturing log C and their version of a wide gamut.
The biggest difference between just using a lut you put directly onto sony slog and this method is that this method will allow you to grade in Arri log C gamma and color space which is a lot easier than being forced to grade in slog with the alternative lut.
Adding: it may be subjective to say grading in one color space and gamma is easier than another but I have yet to meet anyone who preferred grading slog to grading arri log C.
Tony- thank you so much for this! Quick question- do you happen to know the difference / accuracy between this method and something like LUTCalc or Cinematch?
I do not use either of those other options unfortunately so I do not know. You can test the accuracy by shooting a color card though. If you want more accurate you can use ACES however I think The Arri Lut gives a nice initial rec709 look with a good shoulder and allows you to work in Log c instead of slog.
@@TonyDae Thanks! As a Dragon-X owner, do you feel I would be best suited to transform to RedWide/3G10 or go the other way and transform Red to Sony? I know R3D has more latitude for adjustment but in most scenarios RED will be A camera and I also do not love Slog3, so then I would have to transform back from SLOG on both to another space like Arri- which is convoluted, right?
@@thedeltreedotcom you could switch to working in ACES for both of you want to color match rather easily, only caveat is having to work in ACES which is different from Red wide rgb/log3g10.
I guess the big question is what you intend to do for your full pipeline. Are you using red ipp2? If so, you could work in technically any colorspace you want since resolve color management will allow you to do that and still be able to output through red ipp2. If you want ACES, you'll have to conform all footage to the ACES pipeline so you don't have much of a choice but to work in ACEScc(t).
Either way, if you plan to mix and match footage, using LUTs in my opinion wouldn't be the best idea for that. Color management tends to provide more accurate color transforms and tends to make it easier to mix and match. LUTs, not so much.
@@TonyDae Thanks for the detailed reply! I would love to work in IPP2/Redwide/log3g10 - so I think transforming from SLOG3 & .cine would be the way to go right? I need to educated myself on ACES but really I am just far more used to the Redwide color space and R3d's log so its a better starting point for me I think. In addition, I think I might just compare CineMatch to the color transform native in Resolve. They claim theirs are much more accurate- I would like to think their engineers working on it believe in the product so it seems to be worth a shot :)
@@thedeltreedotcom Yes, transforming to Red WIde Gamut RGB/Log3g10 in the node tree then using the red LUTs or using RCM to conform to Red Ipp2 mapping is the way to go. The color management option I think is the better method.
thanks Tony again gd tips
Great video! Where do you think is the technically correct spot to do white balance - pre or post CST? You probably did wb before the CST for the sake of this comparison, but wouldn't be after CST more technically correct? I mean in theory the camera color space could have a totally different white point than the grading space? I assume if we're working with D65 the white balance would be done to that as well, which would be causing errors if the camera profile would work with D50 for instance.
You can do before or after or both if you prefer the results, I couldn't say either is particularly more correct or not. I can't imagine a mathematical transform to inherently be extremely far off if you nailed white balance in camera for instance.
@@TonyDae Thanks for your answer! I did some testing myself which you might find interesting. I found out that the WB location in the note tree can actually have a quite substantial impact in some cases. One of those cases appears when using ACES as intermediate color space, which isn't too uncommon I guess. ACES defines a white point that is different from D65, closer to D60. But most common acquisition color spaces like S-Gamut or Rec 2020 or Alexa Wide Gamut use D65 just like Rec 709 does. So, let's say we're recording Rec 2020 and nailed our WB in camera. Putting this footage into a Rec 709 timeline will show perfect WB as both gamuts share the same D65 white point. But after converting that to ACES, Resolve will show our neutrals as being too cold and slightly to magenta. This is due to the D65 vs D60'ish discrepancy, as our footage is now interpreted as if white should hit the ACES white point. Doing white balance now will obviously cause a warm, slightly greenish cast after converting back to Rec709 - this cast actually kind of reminds me of what I've seen in many movies lately...
Now this shouldn't be a problem when we're working exclusively in gamuts that use the D65 white point. But nevertheless I think it's good practice to do white balance after converting to the delivery gamut to avoid any potential white point discrepancies.
My understanding is that the cst is expecting a particular balance including ACES so I'll get it as close to correct in its own colorspace as possible first, as if the wb was set properly in camera. Generally I'll get wb set before cst and add a correction just after the cst as well. Good research! Link a video to your findings!
@@TonyDae Yes, but that only works if camera and timeline refer to the same white point. You could set up a little experiment: Transform perfectly white balanced footage to ACES and render that. Import that into a standard Rec 709 project and do WB before CST. You will get a warm slightly greenish look after CST to Rec 709.
The same would happen if your camera would work with a different white point than your timeline. Perfect white can have quite different temperatures and tints in different gamuts.
I think we're talking past each other here... Setting your color balance before a transform isn't going to hurt anything and generally corrections will be done post transform as well. Correcting before a transform makes your file as if you white balanced properly when shot, which in my experience gets you closer to a correct transform than not. This will not have any detrimental effects in my understanding.
5:39 the gamma should be “Gamma 2.4” NOT “Rec 709”
Rec 709 is not a gamma
Also if you have your timeline set up to the default values, there’s no need to set the output color space and gamma. Assuming you want to use them
Could be, but I chose rec709. Curve response is more predictable.
Looks like you added more after I responded so here goes...
Yes rec709 is a gamma (or it would not be selectable in gamma), it's a scene gamma in bt1886 and responds more correctly with RUclips/vimeo etc. on output. There's a lot of documentation regarding the difference between rec709 and g2.2 or 2.4, with 2.2 and 2.4 rendering incorrectly depending on devices or through re-encoding via streaming since it expects rec709 scene.
Yes I know out put settings etc with color managed workflows but this isn't a workflow necessarily targeted at only using RCM with specific settings etc. Timeline color space can be different from your output color space when grading in yrgb, so showing output settings is actually important.
what camera do you really prefer in this price range? in terms of color and image quality . a7siii or s1h or bmpcc6k or another suggestion maybe? thanks
It depends on what you're going to do with the camera.
I think that if you want raw and will rig out your camera for cinema production, the bmpcc6k is really hard to beat. Just know that you will have worse battery life and reliability compared to the Sony or Panasonic.
The Sony is, I think, near perfect for the majority of productions I expect from the mid to low end in video production and even has cinema camera functionality if you use an external monitor for levels and monitoring luts. The autofocus is industry leading and the ibis works very well for handheld shots, much better than most cameras for video. The a7siii is really only missing 3dlut installation, false color/waveforms/rgb parade, and internal ND filters to fall short of perfection. The color out of it when converting slog3 via ACES is perfect and the camera can be used in pretty much any situation with great results. I can't see much fault with the camera at all. Also the Sony e mount system has a lot of really good glass that can also be used on their stills camera or cinema camera line.
The Panasonic I think is an in-between camera. It's basically like the Sony except less native glass options and the auto focus isn't nearly as good as the Sony. I think Panasonic needs several iterations before they catch up to Sony bit I wouldn't complain if I had to use the Panasonic or the Sony, both are good.
If you pick Sony or Panasonic and buy Sigma DG DN lenses, you can pay to swap the mount between L and E, so you'll be a bit safer if you decided to swap systems later on.
In terms of color, all of them look pretty much the same if you convert them appropriately through color management via RCM or ACES. The only real difference in my experience is that the Sony and Panasonic tends to be more technically accurate than the bmpcc6k and the Gamut clipping points when using ACES are different in the systems but Gamut limiters and the ACES gamut mappinng dctl fixes any of these issues.
Oh yeah, and if you want to produce HDR content, the bmpcc6k will be better for that using it's internal 12 bit raw compared to the internal recording of either the Sony or Panasonic. In order to really produce good hdr without color posturization and other issues, you really want to start the grade with 12 bit color since it will be mapped to 10 bit bt2020. The other cameras might be okay for HDR, but a 12 bit source will provide enough room to avoid artifacts or other issues compared to 10 bit.
@@TonyDae wow didn't expect this detailed explanation. thank you very much for taking the time to answer . i need to learn about the ACES workflow i just seen your ACES video and i realised i need to dig deep and truly understand it also learn more about color management . thank you very much for your tutorials it's hard to find good information on youtube it's mostly orange and teal looks . you and juan melara are very good.
Np, and don't knock teal and orange, it's highly marketable!
@@TonyDae true
Can you still use CST to convert arri log c to rec 709 instead of lut?
Yes
like it
Is it possible to apply Arri Log C to Fuji?
Yes, use rec2020/flog for input color space and gamma instead of slog.
@@TonyDae Nice.Going to try it later,thanks!