I recently changed to a Fuji X-T3 and am shooting 25p UHD video in F-Log, HEVC 420 10-bit at 200mbps. Cin-GG doesn't seem able to fully cope with it. Let me try and explain.
Cin settings are Preferences/Appearance BT709, JPEG. If I import 3840 x 2160 10-bit and apply Videoscope the first thing I find is, if I set Format to RGBA (FLOAT or 8 Bit) the waveform monitor shows highlights clamped at 100 and they look severely burned out. If I set Format to YUVA-8 Bit only the highest peaks hit 100 with little clamping and only the very brightest highlights are burnt out. It doesn't matter which CC effects I use, it still happens to some extent.
Think fluffy clouds and the myriad of shades from brilliant white to relatively dark grey and I am missing the gradation in the higher shades. This does not happen in, for example, Shotcut or Kdenlive.
There are 3 Fuji LUTs one of which is F-Log to BT709 and if I use that it exascurbates the highlight problem a little. I can achieve a reasonably good highlights correction (in YUVA-8 Bit) with or without the LUT if I use Brightness/Contrast to pull the maximum level down to 95% then use either Colour 3-Way or Histogram and Hue Saturation. If I achieve decent highlights, I get noise in some shadows. Again it does not happen in the other NLEs. It looks the same whether in the Compositor window or in the player when rendered in ProRes.
I wanted you to know because, unless I am doing something wrong, there appears to be a mismatch in Cin.
I don't think I can help you, but maybe my words will give you the inspiration for other tests:
1- Check with mediainfo or VLC (ctrl+I) if your file has the color space BT2020; I noticed that some 4k files have it. If the case, try setting it instead of BT709, but I don't think it will change anything. To aggravate the situation could be the lack of color mangement in CinGG (Shotcut and Kdenlive have it? I don't know).
2- When using an F-Log you must necessarily use the LUT specification or a manual gamma correction (not Brightness/Contrast). They change the luminance without touching the black and white point just to avoid clipping. Only then can you make color correction.
3- For testing, try using the following video posted by another user. This file doesn't give me the problems you found (and it's BT2020):
https://www.cinelerra-gg.org/bugtracker/view.php?id=252
4- Finally, the old message from Mag remains valid, saying to set the project in YUV(A) and not RGB if the video signal is in YUV. Otherwise you get darker images. In your case, however, it seems that RGB leads to clipping in the high lights, and I can't explain that at all. But in my opinion there is a problem of passing between very different color spaces.
https://lists.cinelerra-cv.org/pipermail/cinelerra/2018q4/009903.html
Thanks Andrea, your reply was quite helpful. To take it in order
1. I had already done that with Mediainfo but thanks anyway
2. Understood. I found Gamma only increases gamma, not decreases it, which is why I used Brightness/Contrast. Leaving Contrast at 0, moving the brightness slider I was able to watch the WFM output moving up and down as a block. However, adding Histogram Bezier in Linear I lowered the gamma that way. There is little difference in either the WFM traces or the look.
3. That video simply does not have the range mine has, so the anomaly would not be seen.
4. Once again understood, thanks for the link.
I managed an acceptable CC using the stack Histo Bez, F_LUT 3d, Colour Balance, Histogram, Hue Saturation, Unsharp and Scale Ratio to render out 1080p ProRes, a couple of shorter sequences rendered OK, but an almost 22 minute sequence crashed after 2 1/2 hours rendering, having filled 64gB of DDR4 RAM. I don't run a swap partiton. This is something I have noticed about Cin-GG before. The same sequence renders from Shotcut in 45 min 13 sec. So, one step forward but my other foot is stuck in the mud.
I need to do more testing when time permits, but I am about to embark on a project which I want to get into post quickly, so I shall abandon Cinelerra for the time being.
"I wanted you to know because ... there appears to be a mismatch in Cin."
Maybe true, but GoodGuy and I do not have the expertise to definitely corroborate or to modify Cinelerra code accordingly. I believe this same issue was mentioned awhile back and GoodGuy attempted to understand it and studied the code in detail but this did not result in any conclusion. We are computer people rather than video/camera people and especially lack knowledge in the fine points of Color Correction, clamping, and Videoscope usage, etc.
However, there is always the possibility that an expert with programming skills will show up and help out in this area in the future.
Well Phyllis, I still think you do a good job. No NLE is perfect, the paid ones I used in the past were no better IMO.
