I don't mind my GH2 for stills. Looks about the same quality as my 7D for the most part. But the key is the lenses. You need to shoot with something other than the Panny lenses with the GH2... the color rendition and contrast are horrible. Canon's L or some good Nikon lenses look great on the GH2... for still and video.
GH2 Do you own both Canon and Panasonic?
- Thread starter DevonH
- Start date
If photographing still objects maybe. I was recently asked to do some photo/video of a fundraising event, and another time to capture some kids playing football. I have a beautiful set of vintage primes and fast zooms, but they are all manual focus because Panasonic has a lack of decent native glass. My favorite all-around zoom is a Tamron 17-50 F/2.8, but even if I had the Nikon mount to control the iris, I'm still focusing manually and would have missed either the shot or the focus mark more often than not if doing it manually.
You need to be able to jump in and quickly snap, then back into the shadows, especially with live people who aren't interested in holding still for the pose while you adjust the camera. The wider FOV makes lens selection a bit easier as well.
That is why I ALWAYS use a Canon for photography.
So here's the bottom line, in my opinion:
The Canon is great at stills, but just decent for video.
The GH2 is just OK at stills, but great at video.
However, it is my impression that the GH2 is worse at stills than the Canons are for video. You can use a Canon for video, plenty of professionals have and gotten good results, even thought the GH2 is technically better. But that's more than I can say about the other way around- the GH2 *CAN* take decent stills, but not as easily and it requires a lot more work to make it look good than a Canon does for video.
So, in a nutshell, if you had to pick only one of these cameras, I'd say the Canon is a better jack of all trades. But if video is more important to you, and you just take some product stills on the side, the GH2 is pretty awesome. I would never use it in a pro photo gig tho.
Also, there isn't a profile like cinestyle in GH2-land, right?
I think the best color matching will come from styles that match the most in camera to begin with. That's why I've been patiently waiting to see what happens to that "Adventures in the PSE" thread.
There isn't a profile like Cinestyle in GH2-Land, correct.
It's also not needed, IMO. 5DtoRGB takes care of every complaint I had with the picture profiles. Would I like extended options? Sure, but they aren't imperative any longer.
I'm not sure I understand the purpose of 5DtoRGB. I thought it just transcodes into a format of your choice. How does that help get color definition lost in the 4:2:0 recording?
Cosimo Bullo
Veteran
We too. 5d and GH2. But since driftwood, don't shoot much video with the 5d anymore though.
Sold the RED. I miss it... but don't.
http://www.campcomet.com/archives/836 -- Lots of info and example frames there.
Shoot Epic now, and it's awesome, but GH2 is after me heart. 5D is good in tight spaces, big sensor is nice to have. Keep many options, make happy.
Nice, thanks. Bookmarked, going to read it this afternoon (on a mobile device at the moment). BTW- I love that you use that same girl in all your examples!
Rinse, wash, repeat.
Hahaha someone else commented on that. I do need more test victims but she looks great on camera and doesn't get mad at me, so it works out. She's also a DVXuser, acted in, and helped produce Superseeds.
M. Gilden
Canon DSLR Moderator
No- I think its cool that she's your example in all of them. Not only is she cute, but having the same test subject for all of your examples gives us a reference to compare to (skin tones, etc). Its actually more helpful than you may realize. So, kudos on that.
Never thought about that, the skin thing especially. Huh. Good point.
Ok, I took a moment to read it, and I've got to say that I don't think it applies to folks who aren't using Final Cut as their NLE.
I use Adobe Premiere and Sony Vegas (depending on the project), and both can work natively on DSLR footage without transcoding. If I set my workspace to 10-bit color and import the native clip to color correct / grade, would there really be any advantage to this whole 5DtoRGB thing first?
I thought the idea behind Cinestyle was that you lose highlights and shadows with the 4:2:0 color space, and shooting this flat allows you to bake in detail so that when you spread out the spectrum in post, its all there. The downside to it, however, is that the camera's compression adds noise that just gets exaggerated when applying the grade in post. So its not the cleanest, but it does appear to preserve the most color.
It seems like your complaint perhaps was definition lost in converting to ProRes? I don't use intermediate codecs, so I'm kind of lost in how this is a replacement for a flat picture style?
I think it would be super interesting to read a mini-flow about how to use both in the same scene. Looking for advice to match color, etc.
dishan said:
You'd have to ask Thomas Worth, I'll see if I can point him to this to answer the question. Getting a flat image from 5DtoRGB is why I use it, not sure if that changes on PC platforms.
The 5DtoRGB file, however, is not what's coming out of camera. It's definitely flatter than what I've shot. The ProRes file represents the contrasy image that people have issues with.
Appears to, but in reality it doesn't. Once you get to the grade to correct what Cinestyle is doing you lose a lot to noise. It's almost a placebo effect, at least to me. I did use it at first, and now I'm veering away from it after seeing what it does to the image when you post it.
Nah, you don't lose color defintion or detail going to ProRes, although it adds noise. My complaint was the lack of a flat picture profile, and 5DtoRGB + the Shaders (1.8), and the YbCbR conversion to RGB cures that, for me, though. I dunno how other people work, but it is absolutely a replacement/alternative to a flatter picture style. I even suspect anything flatter than that out of camera wouldn't be good.
But, that's just a guess based on working with this stuff for a while.
I'll see if I can get something together. It wouldn't be TOO hard to match them, and more than likely it would take a really well compressed file to notice the differences if you're structuring the shoot in a specific way: cutting from say a 5D Wide to a tight GH2, or vice versa. But, cross covering with a 5D and GH2 is kind of weird...
Getting color to match would primarily come down to first finding a common exposure, then white balancing both to the same grey card in the same space. You could leave the GH2 in Smooth and use Neutral on the 5D and I am pretty sure they'd look similar enough, as far as color goes, any differences outside of that easily correctable in post.
What neutral settings are you running kholi?
I think the T3i will shine during the reception and similar moments, but the GH2 will continue to be my go-to. Now someone hurry up and take my GH1 and FDs off my hands so I can make this happen.
I'll be shooting several commercial spots in the next few weeks, and I will gladly provide a write up.
My baseline at this point is...
Neutral on the T3i, smooth on the GH2, slightly under expose both.
Bring into Neat Video to denoise
5d2RGB
Start with contrast, then move to color. Be mindful of framing.
I think the T3i will shine during the reception and similar moments, but the GH2 will continue to be my go-to. Now someone hurry up and take my GH1 and FDs off my hands so I can make this happen.
I'll be shooting several commercial spots in the next few weeks, and I will gladly provide a write up.
My baseline at this point is...
Neutral on the T3i, smooth on the GH2, slightly under expose both.
Bring into Neat Video to denoise
5d2RGB
Start with contrast, then move to color. Be mindful of framing.
M. Gilden
Canon DSLR Moderator
The thing is, where is it getting additional shadow and highlight information from? It sounds more likely that this process is simply flattening the image, which would normally be throwing out data. However, since you are also enlarging the color space, it doesn't need to throw out those values- it actually has room to keep them now because of the wider dynamic range. It can just suppress it into the new values that didn't exist before (not sure if the way I am describing it makes sense- it does in my head). Correct me if I'm wrong, but doesn't Final Cut work in the color space of the file being imported? Its been a while since I used it, but I recall it having some sort of limitation that the project and source footage be consistent. If trying to combine different camera sources into the same composition for example, everything required being converted into something like pro-res first (which was a real PITA sometimes). This is as opposed to other NLEs, which just allow you to drop source footage of any type onto the same timeline, and work in a sort of "virtual" color space of your choice, conforming on the fly as it renders.
That being the case (unless I'm totally wrong here), I could see how this would make a tremendous difference in grading footage with Final Cut. You don't get the additional details baked in that the technicolor profile did, however that profile also ends up with the additional noise because of that information being compressed. However, in other NLEs, I would imagine the end result being the same as just flattening the color curve in post before applying a grade.
Yes, because you are trying to "cheat" the limitation of 8-bit color space by adding more detail, which you can then normalize in post. The problem is that the H.264 codec and bitrate being used often adds a subtle (or not-so-subtle) amount of compression artifacts. Usually this is less noticeable. However when shooting that flat, you are getting the same compression artifacts that these cameras always exhibit, however you are later pushing up the color curve to return color to its normal values. That will also bring out the additional compression artifacts, resulting in more noise. Again, I'm wondering if this was exaggerated more in Final Cut versus other NLEs just based on how it works.
Did you ever try it in Premiere?
That commenter on your blog is noticing the effect of the gamma filter, but that doesn't mean the footage looks any different after grading in Premiere. A filter added in Premiere or After Effects on a layer before a color grade may yield the exact same results.
I mean, think about it... where is the additional data coming from? What magic could 5DtoRGB possibly be working that couldn't be incorporated into an editor already?
Last edited:
All of your questions are answered by YCbCr conversion to RGB. You're not losing anything, and you cannot flatten an RGB image then grade. That's what causes you to lose information through generations of exports/conversions.
Just head over to Rarevision, literature and explanations are there.
M. Gilden
Canon DSLR Moderator
Just checked it out, and while the literature and explanations sound nice, it all seems to go back to a specific problem with Quicktime (which has a documented gamma shift bug apparently) and even more so when using FCP. It would appear that none of that is applicable to current versions of other NLEs which are optimized to have their own ways to decode and work with DSLR footage in floating 32-bit full color spaces (contrary to what the literature there would imply, I believe they use their own native decoders now).
After googling "does 5DtoRGB really make a difference", I found this thread on PV:
http://www.personal-view.com/talks/discussion/530/5dtorgb-for-gh2-advantages-vs-disadvantages/p1
By page 2, it looks like either you or some doppleganger with your name joined the discussion, but all the interesting stuff for me happens on page 1. Basically, unless you are using FCP or some NLE that relies heavily on QT H.264/AVC decoding, 5DtoRGB does diddley squat. Well, I mean, it DOES something, just not anything that increases quality in any measurable way.
It appears that the users who said it made a positive difference were using FCP, which all makes sense.
I suppose if you have a really old computer that isn't capable of decoding DSLR footage fast enough to use in a modern editor, or perhaps have an older NLE that relies on QT, then 5DtoRGB is a necessary part of the workflow.
That's my .02