I'm struggling trying to color correct footage taken with a Sony 7sIII and Sony FX3. I used a kyno-like light set, with 5500°K tubes. While they produce a bit of green cast, that is something very easy to correct in post. The problem I'm having is that some red and yellow color are appearing very different from reality, specially when compared with photos taken with a sony 7rIII under the same lights. The photos show almost the "real" color while trying to get that same look from the slog 3 is really difficult. I'm using the official sony SGamut3SLog to 709 conversion LUT and I have tried a bunch of others with same bad results. I have tried to do it in Davinci Resolve as well, with same results.
Pablo, the Slog3.mov file has a 4:2:0 color subsampling and was captured in a lossy format (H.264 codec), compared with the raw photo images, which have no color subsampling or reduced color space. Also, the Slog3.mov file metadata for color space is limited range Rec./BT.709 and 8-bits. Add in the tinted ilghting used and that means that some colors could be affected aversely. Did you manually white balance the video on set? Why use log for footage like this anyway (no real need for "extra" dynamic range with a well-lit setting)?
In other words, when you use a lossy, 8-bit, Rec.709, and log format you won't end up with footage that can be graded much. The lighting and possible lack of manual white balance setting on set also doesn't help. This is a tough one...
Hi Dave, I'm sorry for the .mov, I just exported that from fcpx for speedind up the test.
I've replaced the dropbox file with a portion of the original .mp4 10bit 4.2.2 file the cameras does. I don't think you will notice any difference because of that (in terms of color accuracy).
I have set the white balance to 5500 as the tubes are. I know that the most accurate way to do it would it be doing it manually, but in general this minor casts can be easily corrected.
I don't think the problem here is white balance (IMO) but the LUT or the way I'm trying to recover the 709 space.
I shoot slog3 just because it has more dinamic range, and better handling of iso noise (specially overexposing 1 or two stops). While this was a controlled enviroment I was shooting with two cameras, and while they are very similar they always need some matching. Then I prefer slog just in case some white clothing gets too much light. It's very easy to clip that on standard profiles. I'll try to do a more controlled test to check all of this.
Pablo, thanks for the new file. It makes a huge difference (at least on my system).
Yeah, regarding shooting in log, I should have been more specific in noting that 8-bit log files are rarely worth it (shooting in log with the limited number of colors of 8-bit).
While in the first (.mov) file you posted the orange (on the guy) and the red (on the table) were very close to each other, the new file (.mp4) maintains a decent amount of difference. I have attached a screenshot of the Viewer where I have imported the SLOG3.mp4 clip and just added an "Sony S-Log3/S-Gamut3" Camera LUT. The jacket on the table looks a bit more orangey than in the straight-on shots from the RAW photos, but at least there is a clear difference between jacket colors. Further color correction may help things a bit more. Part of the difference between the RAW photo and the video is a different setup. If both the photo and the video clip were of the same exact scene/setup, they'd probably look much closer.
Thanks Dave, that looks a bit better of what I am achieving with the luts I'm trying. At least the two jackets have some separation as you mentioned. And yes, probably some not-controlled ambient light is making a difference compared with the stills setup where I had my lights closer over the clothes.
I'll throw in my 2 cents here. We shoot in Slog3 frequently with a Sony PXW-Z280 and if grading in FCP we apply the S-Gamut3.Cine LUT as a starting point. I believe a key factor to getting good results involves two primary things. 1. exposing the brightness of the image at 61 IRE (which results in a more contrasty result when grading and also recommended for most of the Sony cameras), and 2. Manually white balancing using an 18% Grey Card. You mention "Kyno-like" light set, which I take as not an actual Kyno light, but some other brand with a similar style. Depending on the CRI rating for that light you might have significant differences between the real white balance vs. setting the value at 5500 because that's what the tubes claim. Your log image looks pretty decent, but might be somewhat overexposed. I'm not sure what the exact recommendations for the Sony A7sIII and FX3 are, but I'd expect them to be pretty close. See the attached document for the technical details of S-Gamut3.Cine and S-Gamut3.
I'm struggling trying to color correct footage taken with a Sony 7sIII and Sony FX3...
If both A7SIII and FX3 are white balanced and exposed properly the image should be virtually identical. If they are both on auto WB or both set to a manual K number (which adjusts only warm/cool axis, not magenta/green axis), they can look very different.
They must be custom white balanced on a white or grey card under the actual lighting conditions of the shoot. They should both be exposed at 61% IRE on a 90% reflective white card. You can do that with Zebra 2, set it to 61% and adjust exposure on a calibrated white card in front of the subject. Be careful to avoid any color spill or splash from nearby items on the card.
The resultant K number may read modestly different between the cameras but that is OK. Each one does a WB calibration separately. Ideally the card should fill at least the center 1/3 of the screen, even though Sony mirrorless cameras have a little steerable reticle.
If that is done properly the resultant image should be very close. It's a good idea to shoot a few sec of the white card or better yet a ColorChecker card for use in post.
Attached is a crude attempt to fix the MP4 footage using FCP but if it was shot properly this would not be necessary except for fine tuning.
Thanks Joema, just to clarify that my issue was not matching the two video cameras I used (7sIII and FX3). They are quite similar fortunately. The issue I'm having is about matching the slog converted to 709 with the A7rIII photo still reference. But the main issue here seems to be the ambient fluorescent light that is interfering with my set up. I should have blocked that room lights (but as always happens, there were other people working there that needed that lights. A black foam in my area would have worked well. Then, I'm not sure if Sony science color is so good and able to reproduce reds and oranges with accuracy. I bet that with my old Canon 5dmark II shooting in magic lantern raw the result could have been much more accurate despite the mixed lighting.
And yes, It seems that the tubes of that kyno-like lights were the warm ones (3200K), not daylight ones, I'm sorry for the consufing info I gave. I use both kind of tubes dependending on the setting. So yes, the exif photo camera tells that I shoot the ARW raw images with 4300K (which seems to give an accurate grey so I guess the tubes were actually the warm ones. Then I suppose that I have shoot at 4300k or 3200K with my video camera as well, if not the images would be completely red. But as I'm not sure now, it contributes to the mess. I'll conduct a test as soon as i can and share the results. Thank you!
Yeah, mixed lighting is a killer for getting color fidelity in the end result (without a lot of work). At least you shot in 10-bit in the S-Log3 color space.
So, using your sample clip as a reference, you will need to do a good amount of color correction work, besides just applying a LUT. LUTs are designed to work with a specific exposure level and balanced (single-source color temperature) lighting.
If you are mostly concerned with the color of the red jacket on the table, you could do an overall correction, then add another one and use a draw mask to isolate the jacket. Then, you could add a color mask or other tools to isolate the jacket on the table and adjust its color separately. I did a two-minute grading on the sample clip (see the attachment). I didn't refine the masking to just stay on the jacket (the presenter's hands get a bit of the coloring meant for the jacket).
So, it's possible up to a point to "fix" things...
...the exif photo camera tells that I shoot the ARW raw images with 4300K (which seems to give an accurate grey...Then I suppose that I have shoot at 4300k or 3200K with my video camera as well, if not the images would be completely red. But as I'm not sure now, it contributes to the mess....
This is a frequent issue when debugging a problem using consumer codecs. They just don't store much metadata. To my knowledge it's not possible to inspect a Sony A7SIII MP4 file and deduce the WB setting used. By contrast the MXF container format used by the FX6 has lots of metadata, including WB value used and many other things.
I tried Sony's Catalyst Browse utility and various command-line tools including ExifTool and ffprobe, none of which would show WB data used in your MP4 file.
The A7RIII still is a 14-bit RAW file. It can be totally manipulated in post. The file EXIF data shows WB 4300 and color mode was neutral using Picture Profile 2. However a RAW file has no intrinsic color baked in -- that is determined when the playback software de-mosaics the image and applies some default settings or metadata settings within the RAW or sidecar file.
It appears the A7RIII may have used a WB preset at 4300K which I think zeros the tint adjustment. Ideally you want to take a custom white balance shot on a calibrated card, with both still and video cameras. However with a 14-bit RAW still it matters less since you can adjust that to anything. Even with 10-bit 4:2:2 video that has less adjustability and some WB is baked in. If you were shooting 12-bit ProRes RAW on the FX3 captured to a Ninja V, you could probably adjust the color however you want.
That said, if both still and video cameras were properly white balanced and exposed on a 90% reflective target, they would likely look much more similar. Sony's exposure spec for SLog3 is 61% IRE on a 90% reflective target. You can achieve that using Zebra 2 set to 61%. When the Zebra pattern just starts to appear on a white target in front of the subject, it is exposed properly for SLog3. Then take the WB calibration shot.
When properly exposed and white balanced, the FCP built-in SLog3/SGamut3.cine camera LUT will look excellent.
That said you are talking about matching FX3 SLog3 video to an A7RIII still which does not have SLog3 and is using a baked-in PP2 picture profile. That may not look the same. But I think the main problem is the video cameras were not exposed and white balanced properly.
If you dial in a K number or use a WB preset instead of shooting a white balance target, this will only adjust the warm/cool axis, not the magenta/green axis. Taking a WB shot will adjust both of those.
Even though the Sony mirrorless cameras have a little steerable reticle for taking the WB reading, I personally don't trust that. It is likely better to fill at least the center 1/3 of the frame with the WB target. Be careful about color splash or color contamination on the WB target from adjacent objects or apparel.
The moral of the story to take away from all this is to WB using at a minimum a gray card. No matter what the tubes say (5300) or no matter what the soft box lamp says (4500) and no matter what the fully adjustable LED lights say I'll always WB with an 18% gray card in camera & roll a few secs of the scene with the card in the frame, particularly on a product shot where the colors have to be accurate.
Every Mac comes with Digital Color Meter if you want to dial in exact values.