GH2 So, what's everyone's favorite patch? And, if you didn't hack your GH2, why not?

I tried Sanity5 and hated the look I got when shooting anything with fast motion, very strange images. I have stuck with the stock 1.1 as I am more concerned about reliability as most of my shoots are live events and there has never been an option for a second take!
 
Thanks for the words of support from everyone who's recommended my latest Flow Motion v2 patch for the GH2. This patch took quite some time to develop, as I wanted to optimize it for consistently excellent motion picture quality across the broadest range of NTSC and PAL video modes and shooting conditions.

As was pointed out earlier, Flow Motion v2 is designed and tested to work reliably at 100Mbps peak bitrates on standard 30MB/sec Class 10 SD cards with in-camera playback supported in all AVCHD video modes. In addition, extended-length recordings with 4GB file-spanning are supported on all Class 10 cards in FM2's 60Mbps 24L, FH, and H modes. (For those with 95MB/sec SD cards, file-spanning is supported in all AVCHD video modes except NTSC HBR and FSH modes.)

For those interested in downloading Flow Motion v2, here's a link to the details:

http://www.dvxuser.com/V6/showthread.php?283724-GH2-100Mbps-Flow-Motion-v2-Patch-Banding-amp-Blocking-Be-Gone!&p=1986152355&viewfull=1#post1986152355
 
Last edited:
For the moment I am using my GH2 unhacked with the 1.1 firmware. I always intended not to rush into hacking the camera as I want to see exactly what the camera out of the box can give me. And I have been overall suitably impressed. My only concerns with hacking is what seems to be an increased number of people saying their camera froze or card error messages. Plus the need for the more expensive cards to improve reliability and of course the bigger file sizes. If and when I do give hacking a go I will try Flow Motion as that seems to have mainly positive and few negative comments against it.
 
I have Sanity 5 Loaded. I wanted a patch that would give me a quality boost with the 60p footage I was shooting, the 24p stuff looks great too. It has less macroblocking in the shadows and grades a lot nicer than stock settings. Works pretty flawlessly with our transcend class 10 cards. We are 70% finished shooting a 'found footage' type horror film with that patch. http://www.facebook.com/pages/The-Changing-of-Ben-Moore/180011858692775

Flow Motion looks inviting. I definitely give it a try

https://vimeo.com/39384879#at=0
 
Flow Motion has been the jewel for my GH2. I've tried almost every hack and Flow Motion gives me the best motion (panning/tilting-fast/slow) video images. I've shot with a variety of legacy lenses and the video looks...well not like video.
 
Obviously we pick a favorite based on 1) what works on our equipment and cards 2) what looks good to our eyes 3) etc.

Has anyone ever done a comparative shootout in a controlled environment with the major different hack versions? I mean something like what you see in the FXPHD videos when they compare the 5D M2 against the M3. They have a well-lit moving Golden Retriever in the shot with a highly detailed sun shade reflector from a car and a green screen behind the dog so there was a ton of high frequency data as well as fine detail to compare and look for macro blocking, etc. With the right kind of angular lighting and textures we could probably also include a banding section and moire test.

I'd love to see something organized like that. I'm not currently equipped to pull that off right now, but I might be within a few months. It would just be nice if there was something a bit more concrete and measurable than subjective opinion when choosing a hack. I'd love it if I knew hands down that hack X will give me better details and more manageable noise in low light footage compared to hack Y, but hack Z is best for outdoor natural environments that are well lit, etc.
 
In the end it is subjective, isn't it? Technical pixel peeping means very little to each of us as individuals in such an artistic endeavor. In the end you try them for your video making; others test may not satisfy you. I have tried a few and have settled on Flow Motion 2.02 The only troubles I have had are very predictable, namely: highly detailed subjects that are changing with 720p60 (you need to test to learn what this is) which means I must use H setting when I want to film this in 720p60. All other settings I use are solid. My subjects are family and I am aware when a video clip can't be re-done and set my camera accordingly.
 
Sure, but there are some elements to the hacks that can be less subjectively analyzed, like the examples I gave. Some hacks are designed for better low light, or better motion. How well they really do against other hacks really isn't that known. Two people can look at the same video and feel different about the quality. Sometimes it just feels like we're splitting hairs. I'm settled on Flow Motion because it works reliably with my configuration and cards and I think it looks great, but maybe I'd switch the hack to a different flavor for specific types of shots in specific conditions if I knew there could be a real gain there.
 
With DSLR video cameras, my experience is that subjective image quality is far less an issue than eliminating telltale digitizing artifacts and flaws. H.264/AVCHD encoders were not originally optimized for image capture, they were designed for highly-compressed, consumer-grade content delivery, such as Blu-ray movies. The encoding compromises necessary to cram feature-length films into BD-discs are transient and hard to spot on an HDTV. But when capturing source footage for post-production, minute technical flaws can ruin the gradability of a shot and require a great deal of work to salvage.

The Panasonic GH2 is unique in employing an AVCHD encoder that is implemented completely in software, with no fixed-bitrate hardware components. That is what makes it supremely hackable, but it also restricts the encoder to using a simplified subset of H.264 encoding techniques. As a result, the unhacked GH2 is prone to producing coarser macroblock artifacts than competing DSLR's (as well as the AF100) that use hardware-assisted AVCHD encoders.

Fortunately, PTool allows us to boost the GH2's bitrate to over 100Mbps and customize its quantization tables to compensate for its unhacked shortcomings and virtually eliminate perceptible digitizing flaws. For panning and shots with rapid motion, this enables the GH2 to produce smoother motion picture quality than any other H.264-based camera that I've seen. What matters in this case is not so much static image quality, but the perceptually flawless details of its encoding process.

For those interested in examining sample files, I've uploaded original GH2 footage comparisons in the Flow Motion v2 link in my signature below:
 
You know, Driftwood's Cluster 6 is also pretty frickin sweet for standard gop size. I think that patch and flow-mo are my favs for reliable, non-very large file size hacks. Flowmo i think does better on motion than cluster 6 (naturally given the lower gop size), but cluster does better on detail. But looking forward to the cbrandin matrix on canis and, hopefully, it'll get transfered to lower bitrate hacks like cluster.
 
BTW, an easy test to confirm how well competing hacks do on macroblocking and fine detail is to shoot at night with a controlled key light. Set up a scene, and shoot the same thing with different hacks. In this way, I can see that driftwood hacks like cluster 6 (and surely Sedna, Canis, etc), do better on fine detail rendering than Flowmo.

That's **one** consideration **for me** in choosing hacks. I think we'll never get to a point where everyone uses one hack or recognizes one hack as THE BEST. And it shouldn't be that way.

What's objective in hack comparisons is how well different hacks do on different dimensions such as:

- fine detail
- motion fidelity (really hard to test)
- file size (effeciency)
- in camera playback
- cost of sd cards
- (color ? -- not sure if matrix adjustments affect this)
and so on.

Those things are objective--and I wish we could devise some objective test for motion fidelity. But it is really key to realize that **how important** each of those things is to YOU is subjective. So, ultimately, the choice is yours. Happy shooting! :)
 
Actually, I found flaws in motion fidelity to be one of the most readily visible artifacts in unhacked GH2 footage. For this purpose, flowing water is an excellent test case, as it combines smooth contours with rapidly varying independent motion. It also demonstrates an important principle about motion pictures. Very often, it is not necessary or even desirable to record the precise details of a textured object. What the eyes find most pleasing is a realistic rendition of objects in motion, with clean edges free of distracting digital artifacts. In the past, a fine layer of noise was often relied on to conceal a wide variety of flaws and technical shortcomings. I prefer to start with a clean slate, as flawless as the technology can produce. If desired, a controlled level of noise can always be layered in post.
 
lpowell said:
Actually, I found flaws in motion fidelity to be one of the most readily visible artifacts in unhacked GH2 footage. For this purpose, flowing water is an excellent test case, as it combines smooth contours with rapidly varying independent motion. It also demonstrates an important principle about motion pictures. Very often, it is not necessary or even desirable to record the precise details of a textured object. What the eyes find most pleasing is a realistic rendition of objects in motion, with clean edges free of distracting digital artifacts. In the past, a fine layer of noise was often relied on to conceal a wide variety of flaws and technical shortcomings. I prefer to start with a clean slate, as flawless as the technology can produce. If desired, a controlled level of noise can always be layered in post.
Click to expand...

Thanks for mentioning this.
I SO agree - I feel like I just can't watch anymore digital footage with weird edge or cadence artifacts during motion. Higher levels of detail are sometimes needed. But natural motion rendition is ALWAYS needed. (Of course I'm not talking about special effect frame rates or motion special effects.)

Sometimes I REALLY miss my DVX100 because those issues were not...well...ISSUES.
 
You must log in or register to reply here.
Back
Top