Dear Community,
Is there any way to deal with Cine-D footage on Shotcut?
I’m planning to buy a Lumix G80 camera and start playing with log footage.
Shotcut is my editing baby.
Thank you.
Kind regards,
J.
Yes, one option is to get a Cine-D to Rec. 709 LUT, possibly from Panasonic but there might be third party ones that some prefer. Also, since the advantage of log is greater flexibility of color grading you also want to use the Color Grading or Levels video filters. However, since Shotcut’s CPU effects engine is 8-bit integer I think you should place the LUT filter after these filters. Otherwise, the LUT might cause clipping that you were trying to avoid by using log.
The GPU Effects mode uses 16-bit floating point but it currently lacks a LUT filter. So, attempting to use that combination can result in some clipping or quantization loss (information slightly inaccurate due to the low precision of 8-bit numbers).
Another option is to shoot some same things in both Rec. 709 and log, and then use Color Grading filter without a LUT to develop your own base preset. The goal is to have a preset that you apply to your log footage to bring it very close to Rec. 709. Then, for each shot you apply the preset and fine tune it for that exposure. That way you avoid 8-bit integer degradation between successive filters. This works both with and without GPU Effects, but with GPU Effects and only using GPU filters you can read and export 10-bit without any degradation.
Thanks for your reply.
You have shared very helpful information!
The base preset is definitely the way to go and I will share the results with the community next year.
Out of curiosity, upgrading CPU effects engine to 16-bit integer is it something that will be considered in the future?
Cheers,
J.
I have a G85 (the North American version of the G80) and love it. I’ve included some sample “log” footage and notes below.
I put “log” in quotes because Cine-D and Cine-V are not true log formats due to the G80/G85 being an 8-bit camera. Trying to map the entire sensor response equally per-stop to an 8-bit curve would cause horrible posterization or banding. Therefore, what Cine-D and Cine-V do is keep the color primaries (or at least axes depending on settings) essentially the same as sRGB/Rec.709, then moderately tweak the gamma curves to record more dynamic range than usual.
When I use 8-bit cameras and codecs, I try to make the in-camera shot be as close to the intended final look as possible so that minimal post-production is necessary. Any stretching on 8-bit footage risks banding pretty quickly. With 8-bit “log”, there is stretching required just to get to a normal start point, and then the creative grading process stretches things even further. The footage may or may not hold up to that much stretching. So I try to nail my final look in-camera to the best of my ability. This is pretty much the opposite of 10-bit workflows, where some people just “get it in the can” without trying too hard, then enjoy the flexibility of stretching it all over the place in post without consequence. Lucky them.
To help nail the look in-camera, there are five picture profiles I frequently use, in this order of frequency:
- Natural: It does not claim to be log, but it is still quite flat. It is particularly useful for forest interior. It is also low saturation, which is great to prevent vibrant flower colors from blowing out even in shaded areas. Grading is easy because brightness is designed to look normal, and it’s usually just the saturation that needs optimization.
- Cinelike-V: This pseudo-log format slightly compresses the range so that more highlights can fit in, while keeping the contrast in the midtones pretty punchy. For high-dynamic range scenes where you want minimal grading to do, this profile is great and is often immediately usable.
- Portrait: As the name suggests, it can save some grading time by optimizing for skin tones. It gives more tonality to skin tone colors, which prevents patchiness and blockiness during close-up shots.
- Vivid: If I’m outdoors in winter forest where dead grass and light tree bark look bland and colorless, then I can use Vivid to boost those colors in-camera enough that I can skip grading in post (or at least have a better start point).
- Cinelike-D: Like Cinelike-V, it compresses the range even further to maximize captured dynamic range (within reason for 8 bits). However, it sacrifices contrast to do this, so it requires additional grading effort.
Where Cine-D and Cine-V come in particularly handy is for high dynamic range scenes, like a shot of someone’s face when they are backlit by strong sunlight. I can expose more for the sunlight to prevent some blowout, and the flat nature of “log” will super-boost a shadowy face into visible usefulness. I know this is not how log is intended to be used, but hacks are a necessity when the whole pipeline is 8-bit. It works, that’s all that matters to me lol. Anything that optimizes the tones and gets them close to their final ranges using the 12-bit ADC values from the sensor will make the footage look its best before being reduced and encoded to 8-bit, where a lot of data will be permanently lost. The G80 in particular has a tool called Highlight/Shadow which allows the tone curve to be customized before encoding. This is such an important feature, particularly for boosting shadows, that I assigned it to the Fn11 button (the little push-button inside the thumb dial) and use it all the time to optimize for each scene. Here is an example of it in action:
All that to say, if a scene doesn’t have high dynamic range, then I will avoid Cine-D and Cine-V because they will burn valuable bit-depth trying to capture a wider range of brightness than I need. I will instead use Natural, Portrait, or Vivid if the scene’s dynamic range fits comfortably on the histogram without clipping. If clipping is an issue, then Cine-D and Cine-V become hack workarounds. You may have heard that people with GH5 cameras will set Contrast and Saturation to -5 in order to make the captured range extend even further. However, the GH5 is a 10-bit camera, so they can get away with that. Taking G80 footage with -5 settings and grading it back to normal in post is a recipe for banding unless the dynamic range of the scene actually used the full histogram, and you specifically want the flatter look.
If it does come to grading Cine-D/V footage, note that the gamma curve will act inconsistent because it’s a wider range. This means that the three brightness controls provided by the Color Grading filter will not be enough by themselves to manipulate the footage. You’ll also need a Contrast filter to spread or contract the midtones (depending on the scene) in conjunction with Color Grading, because the range compression in Cine-D is different than Rec.709. The RGB parade scope will make it obvious where the midtones are bunching up. By separating them with Contrast, they can be better controlled by Shadows and Midtones in Color Grading. Also, the G80 has a tendency for red tones to run away to full saturation earlier than other colors, which can look particularly bad on people with a sunburn (don’t ask how I know this lol). A way to fix this is to use the “Old Film: Technocolor” filter, set both sliders to 100 (the do-nothing value), then adjust the Red slider to reign in the reds. The result is more natural than dropping red in the Color Grading filter.
Hope that helps. Now for some sample footage:
Hi Austin,
Thanks for sharing your experience with the Lumix G85.
That will help me to narrow down my tests with the camera.
Also it is good know that G85 is a 8-bit camera which will fit nicely with Shotcut engine.
Cheers,
J.
This topic was automatically closed after 90 days. New replies are no longer allowed.