GH2 Incoming Canon shooter: I think I want a GH2, but I have a question.

H

HansSteinert

Well-known member
So the resolving power of the GH2 along with the bitrate hack blows my mind.

What kind of latitude flexibility can I get with Gh2 picture profiles compared to Canon cameras? Because if I am forced to play with an ultra contrasty image, I won't exactly be happy. Is there a relatively flat and digitalsharpness-free solution for the GH2? Not Technicolor, even I think that's overboard... but something greater than Canon Neutral with contrast and sharpness dialed all the way down?
 
It depends a lot on the lens. If you plan to use the m4/3 lenses, they will look a lot different from legacy lenses. It is my opinion that the most pleasing images from GH2 come with classic lenses (and new but still classic lenses). As I am always quick to provide examples, here is one filmed with the GH2 in factory configuration with no customized settings using a Minolta MD 135 mm lens: http://www.fourseasonshd.com/cgi-bin/csvsearch.pl?db=db2&tp=tp3&search=1131

Another note is that I am not sure if the gamma curves issue is overestimated, because no contrast manipulation in the camera will affect the sensor output. So using the most linear setting the camera offers should be a good basis for 32-bit grading, where you can manipulate contrast and color to your liking. Of course it is a bit different if one needs to bake in suitable contrast, gamma and colors in the camera.

Here is probably the first scene I recorded with a GH2 last summer, I only pulled the camera body out of the box and installed the only fitting and very modest legacy lens I had at the moment and fired away: http://www.fourseasonshd.com/cgi-bin/csvsearch.pl?db=db2&tp=tp3&search=1112
 
Last edited:
1. Set LCD and Viewfinder up for exposing: dial down the LCD contrast slider (To the left) and up the brightness (to the top) in the settings. Then, switch LCD mode to *A1

2. Use Cinema, Standard or Smooth profiles. Standard is the most linear, Cinema looks the best but pushes the entire range down a fair amount (not that great for lower lit situations), smooth is the noisiest to me but opens up the curve.

3. Transcode through 5DtoRGB Batch > 709 Full Range 4444 or DPX for grading, etc.

Only way to work with the GH2 Hack for a strong finish IMO.

can see more here: Mini-Flow Part 1 | 5DtoRGB makes your footage prettier
 
Kholi's given a great, concise write-up. I agree with his filmmodes description, and would only add that nostalgic technically has the most dynamic range (about 1/3 stop more than smooth). But nostalgic also has a pretty pronounced yellow color-shift, which for me makes it useful only in very narrow, if any, contexts. I've got to double check, but if I remember this affects it even with a white balance to grey card.

I've been out of the loop for a while, but I REALLY need to try 5d2RGB (especially since I started a thread over at personal-view asking about its value!). Also look forward to trying Kholi's LCD tip!
 
Has anybody tried levels on a image filmed with standard film mode and compared to image filmed with a flatter film mode? I'm betting the results will look quite similar. I don't think these cameras have sophisticated electronics to manipulate high bit depth curves before laying down to 8-bit.
 
I put the two images in Kholi's link on top of each other in Adobe Photoshop and by using levels and curves was able to make them identical. So the images appear to contain the same image information, just with different values to start with. If this is the case, then there is no difference which version one chooses as the basis for grading in 32-bit.

But the images were probably scaled down versions, so comparing the original version would only tell whether they are the same information only with different values.
 
The GH2 doesn't have as much latitude as some other cameras, but it's enough to make beautiful pictures. Take care to nail the exposure and you'll be fine.
 
Kholi said:
1. Set LCD and Viewfinder up for exposing: dial down the LCD contrast slider (To the left) and up the brightness (to the top) in the settings. Then, switch LCD mode to *A1

2. Use Cinema, Standard or Smooth profiles. Standard is the most linear, Cinema looks the best but pushes the entire range down a fair amount (not that great for lower lit situations), smooth is the noisiest to me but opens up the curve.

3. Transcode through 5DtoRGB Batch > 709 Full Range 4444 or DPX for grading, etc.

Only way to work with the GH2 Hack for a strong finish IMO.

can see more here: Mini-Flow Part 1 | 5DtoRGB makes your footage prettier
Click to expand...

Thank you sir. How much grading flexibility is there with Cinema and Smooth given the route you listed above?

mindblown at how good the flat conversion is.

Anyone wanna shoot me a raw random clip shot from a hacked GH2 for me to play with? I want to play with this technique and try grading it a bit.
 
Last edited:
Aki_Hartikainen said:
I put the two images in Kholi's link on top of each other in Adobe Photoshop and by using levels and curves was able to make them identical. So the images appear to contain the same image information, just with different values to start with. If this is the case, then there is no difference which version one chooses as the basis for grading in 32-bit.

But the images were probably scaled down versions, so comparing the original version would only tell whether they are the same information only with different values.
Click to expand...

Actually, in practice, there is definitely a difference in the way 5DtoRGB (if that's what we're talking about) spits out files and how final cut handles them.

If we're talking about film modes, again in practice, you want to give the final AVCHD 4:2:0 as much data as possible. That means taking into consideration how the film modes expose mids to lower mids. Try it if you've got the time: shoot cinema in a somewhat decent lit place then shoot smooth without changing anything, shove the curve back down to match cinema.

It's slightly cleaner, the results.

In my opinion (take it or leave it!) what you do not want to do is put an image to the card that has under-exposed mids unless it's going to stay that way.

For the most flexibility in post:

1. DO NOT FLATTEN THE IMAGE LIKE CRAZY. This is a supreme no no. My Cinema settings are +2 on contrast, -2 on sharpening, 0 on saturation, -2 on Noise.

When 5DtoRGB gets ahold of that it's JUST flat enough to push around without dissolving into a mushy mess.

2. Overexpose mids slightly when balancing light, record them, then push them back down later. If I had to choose between a sky and people, I'm gonna choose people on this camera. Get yourself a polarizer or a grad .6 ND soft if you want skies.

In fact, get a Polarizer anyway, it helps a ton when dealing with highlights on just about any surface, including skin.

If you plan on moving the image around, this is what I've experienced and what I'm sticking to.

Naturally, the BEST result you're gonna get out of sub 15K cameras is shooting as close to your look IN camera as possible. It was true with the DVX100, it's still true today.
 
Last edited:
I'm not sure where the windows version is at right now. I'm on Mac. I need to investigate that when I get a bit.
 
I researched it and yes, it is Mac only. If I convert to uncompressed 4:2:2, couldn't I just do the gamma change myself in post since the image is literally uncompressed rather than prores?
 
Kholi said:
Actually, in practice, there is definitely a difference in the way 5DtoRGB (if that's what we're talking about) spits out files and how final cut handles them.

If we're talking about film modes, again in practice, you want to give the final AVCHD 4:2:0 as much data as possible. That means taking into consideration how the film modes expose mids to lower mids. Try it if you've got the time: shoot cinema in a somewhat decent lit place then shoot smooth without changing anything, shove the curve back down to match cinema.

It's slightly cleaner, the results.

In my opinion (take it or leave it!) what you do not want to do is put an image to the card that has under-exposed mids unless it's going to stay that way.

For the most flexibility in post:

1. DO NOT FLATTEN THE IMAGE LIKE CRAZY. This is a supreme no no. My Cinema settings are +2 on contrast, -2 on sharpening, 0 on saturation, -2 on Noise.

When 5DtoRGB gets ahold of that it's JUST flat enough to push around without dissolving into a mushy mess.

2. Overexpose mids slightly when balancing light, record them, then push them back down later. If I had to choose between a sky and people, I'm gonna choose people on this camera. Get yourself a polarizer or a grad .6 ND soft if you want skies.

In fact, get a Polarizer anyway, it helps a ton when dealing with highlights on just about any surface, including skin.

If you plan on moving the image around, this is what I've experienced and what I'm sticking to.

Naturally, the BEST result you're gonna get out of sub 15K cameras is shooting as close to your look IN camera as possible. It was true with the DVX100, it's still true today.
Click to expand...

Thanks a lot for your well written posts and 5DtoRGB tutorial Kholi, that's really going to help a ton getting the maximum color grading quality out of my little GH2!

This thread should be made a sticky!
 
No problem.

Also, when I say "slightly overexpose" I'm saying that what you would normally let go slightly under, expose for that. Shadows will come up a bit, highs definitely, but if you've ever tried to open up an underexposed image (By open up I mean hoist the mids) you'll see why.

If you have to, use ISO to do it. Just don't go above 1600. LPOwell (I think it was he who stated) seems to be right in that anything about 1600 ISO is software. Can't remember exactly how he put it but the threshold is definitely 1600.

I encourage people to play with those three film modes and see what you like better for different situations.
 
Hey kholi I'm wondering if the 5DtoRGB conversion is necessary if I'm not working in Final Cut Pro.

I'm working with Premiere Pro Cs5.5 and it can handle the mts files natively.

However, with the conversion, will I be getting a less contrasty and a flatter image?
 
I have used Cineform intermediate 4:2:2 AVI codec for several years with good results for the reason of getting upsampled colors, plus GoPro Cinefoerm has a fast 32-bit primary color corrector First Light. However, that gamma most likely manipulates existing values, so any decent color corrector can provide that result.

I am studying entirely different approach for the most gradable GH1 and GH2 material (examples below)

GH1purpleFilter.png


GH1colorsAdjusted.png


It can be said that the red and blue channels are crippled in these cameras, and no grading will fix that. However, using a suitable filter on the lens provides oversaturation of certain hues, which can be later compacted in post thus minimizing mosaic. If I tried to get the reddish version without lens filtering, mostly just mosaic would show up, and this effect works to the other direction too.

The lens filtering or lighting coloring has to be done so that compacting the troublesome channels lands one to the desired end colors for maximum digital noise reduction of rather poor 4:2:0 material.
 
That's pretty interesting. Do you use something like this? Do you have any before/after shots that show a scene with some details?
 
Work with the lens filters is very preliminary at this point. For purple I have used a low cost chinese round filter and for blue I have an old Canon CCB4 round filter. The GH1 has distinct vertical striping visible in low light scenes, and with the blue filter I have managed to get rid of that. Some GH1's have more of the vertical striping than others, but none is completely without it.

If you are really interested seeing the effect, I have one video on YT that shows the problem: http://www.youtube.com/watch?v=1QycpJp-YGA

You will need to go to 52 seconds in the video to see a car indoor scene of a shifter. Right to the shifter are the GH1 vertical stripes, and I already used all sorts of things to lessen their visibility. But in this video shot through the Canon CCB4 filter the image is free of the vertical striping and quality is rather good: http://www.youtube.com/watch?v=bnw5jXCHG4k

The science behind this is that by over-saturating the problematic hues already on the camera sensor one is actually filming with a larger bit depth than 8-bits for that particular range of values. When later compacted in the post the values become closer to each other, and this is the same effect as larger bit depth would have. But only for a limited range of values. So it's not a solution without it's own limitations. However, it has allowed me to do low light scenes with GH1 that otherwise would have been ruined by the vertical stripes. Ideally there would be standard settings with the filters for particular situations.

Hopefully I could create a better demo soon.
 
How much of a difference does 5DtoRGB make? I remember downloading a trial and I really disliked the results I got with my 7D footage, whereas the MPEGstreamclip footage looked great. I've also seen tests where people compare the two, and the MPEGstreamclip samples look much better to me. So why all the praise?

Also, I'm about to get my GH2 and was wondering what codec I should use? I always used ProRes 422 LT with DSLR stuff because anything more is detrimental to the footage; would I be safe to assume the GH2 is best suited to ProRes 422 LT as well, even with the 44mb hack?
 
Lens filtering can produce more gradable and less posterized footage. Below two samples that were filmed with and without a 1.5x blue filter (0.5 f-stop attenuation):

blueLight1.png


blueLight2.png


The filter ate 0.5 f-stops, so to compensate I opened the lens up one full f-stop. Later in Photoshop I lowered exposure 0.5 f-stops. This created flatter and less posterized image (more values to begin with), thus better candidate to begin grading.
 
You must log in or register to reply here.
Back
Top