GH4 Who'll be first to post up SHOGUN footage??
- Thread starter wwjd
- Start date
dustylense
Veteran
I'm kinda on the fence whether I "need" the Shogun or not. Mathematically, we could argue the image is better. Visually? Is it $2000 better? But I hear the monitor is pretty darn nice and I've spent $1200-1500 on monitors alone. So for a few more Benjamin's you get a 4K recorder. The one thing that has my finger on the order button is my A7s though. Stuck between a rock and a hard place I guess.
Here are a couple ungraded and graded versions I've done with the Gh4 shooting in Cine-D. They are pretty darn flat to start. After the grade I have to sit pretty close to the monitor to see degradation to the clips. Sitting back and taking the image in like a normal human would it's really hard to pick them up. (obviously they are forum compressed but they look great on my 4K monitor).
Here are a couple ungraded and graded versions I've done with the Gh4 shooting in Cine-D. They are pretty darn flat to start. After the grade I have to sit pretty close to the monitor to see degradation to the clips. Sitting back and taking the image in like a normal human would it's really hard to pick them up. (obviously they are forum compressed but they look great on my 4K monitor).
Attachments
I agree, Dusty, seeing maybe $300 worth of improvement over the internal recording. I believe the internal is good enough to project at 4K in a 4K theater and look just great. If I was commercial, and each gig paid off more gear, I'd get a recorder in a second. But as more or less an amateur hobbyist, I don't see a $2000 difference, even if the grading is 10% more pushable.
But, downloading those free examples now and will pixel peep the night away. 4K blows me away.... a tiny slight improvement in colors in 4K does not.
But, downloading those free examples now and will pixel peep the night away. 4K blows me away.... a tiny slight improvement in colors in 4K does not.
I downloaded the footage from that NF video above, and did some of my own testing with Vegas Pro 13, at 32bit floating video mode. matched the scene framing, and brightness, looked for banding, detail, and finally quick&dirty grading with my typical chain.
Although this was a completely unscientific test - I'm just a guy, not a pro at anything - I could SEE the differences caused by the internal compression.
- The BANDING was different, kinda not better or worse, just different... really pushing it made the Shogun band seem not as good TO ME.
- Zoomed in, the detail was fuzzier on the internal clip. The slightest bit of sharpening made it SEEM equally detailed as the Shogun clip, and pushing the sharp (or unsharp mask) could make the internal APPEAR sharper, more detailed. In my test, SHARP level 1 (out of 10) was all it took to look just the tiniest pinch sharper than the shogun clip. (both clips had the slightest halo - guessing camera sharp was left at 0 or maybe -1)
- Quick grading was just slightly different - both totally useable. I would not claim one was better or worse looking than the other.
Obviously, the recorder DOES get you higher quality footage in bit rate and color, spec wise.
The ONLY thing this proves to me, is that >>I<< probably don't need Shogun AT THIS TIME, since I am not working commercially. Not trying to talk anyone else out of it... you CAN see the difference in the details.
The pics with the bird and the plane in it are the INTERNAL ones. Note: the sun had moved between the takes creating different shadow locations and lighting.
...and I am sure looking at these on this website compresses these to not showing much difference.
you can download the 4k Vegas Frame grabs here:
https://copy.com/fobVlU7daspATCqg
Although this was a completely unscientific test - I'm just a guy, not a pro at anything - I could SEE the differences caused by the internal compression.
- The BANDING was different, kinda not better or worse, just different... really pushing it made the Shogun band seem not as good TO ME.
- Zoomed in, the detail was fuzzier on the internal clip. The slightest bit of sharpening made it SEEM equally detailed as the Shogun clip, and pushing the sharp (or unsharp mask) could make the internal APPEAR sharper, more detailed. In my test, SHARP level 1 (out of 10) was all it took to look just the tiniest pinch sharper than the shogun clip. (both clips had the slightest halo - guessing camera sharp was left at 0 or maybe -1)
- Quick grading was just slightly different - both totally useable. I would not claim one was better or worse looking than the other.
Obviously, the recorder DOES get you higher quality footage in bit rate and color, spec wise.
The ONLY thing this proves to me, is that >>I<< probably don't need Shogun AT THIS TIME, since I am not working commercially. Not trying to talk anyone else out of it... you CAN see the difference in the details.
The pics with the bird and the plane in it are the INTERNAL ones. Note: the sun had moved between the takes creating different shadow locations and lighting.
...and I am sure looking at these on this website compresses these to not showing much difference.
you can download the 4k Vegas Frame grabs here:
https://copy.com/fobVlU7daspATCqg
dustylense
Veteran
I too played with the same footage and saw some concerning things.
I pushed this frame pretty hard along with secondary keys on the sky and the red rocks:
Applied the same grade to the Atomos PR and the Native Gh4 file.
When zoomed in I see white haloing in the Atomos footage where I don't see it in the native footage
Look around the tree branches.
Atomos:
Native:
Look at the edge of the cliff.
Atomos:
Native:
Sure, the blue banding of the sky is "better", but only by 10-15%. When I play back if I take away the banding test, I actually think the native file performs slightly better. The haloing is just plain weird. Exact same settings on each clip.
I pushed this frame pretty hard along with secondary keys on the sky and the red rocks:
Applied the same grade to the Atomos PR and the Native Gh4 file.
When zoomed in I see white haloing in the Atomos footage where I don't see it in the native footage
Look around the tree branches.
Atomos:
Native:
Look at the edge of the cliff.
Atomos:
Native:
Sure, the blue banding of the sky is "better", but only by 10-15%. When I play back if I take away the banding test, I actually think the native file performs slightly better. The haloing is just plain weird. Exact same settings on each clip.
I love it! So psychedelic! I saw the halos as well. But I am wondering if that is from the GH4 left on SHARP 0, and that Atomos captured it better, where it is lost more in GH4 internal compression? My experience is at SHARP -2 the halos are nearly gone. I wonder if there are SHARP settings in the Shogun as well? We need a little more info from Neumann Films, I think. Regardless, my skirt is not blown up by 10-bit.
Maybe someone will soon post something that IS an amazing display of 10-bit footage over 8-bit. I'm now more impressed with what Panasonic accomplished with their internal compressed file! ...with a 4K camera attached to it as well!
Maybe someone will soon post something that IS an amazing display of 10-bit footage over 8-bit. I'm now more impressed with what Panasonic accomplished with their internal compressed file! ...with a 4K camera attached to it as well!
goodgoings
Well-known member
Can you link to some awesome footage because I'm not seeing any yet. I had high hopes it would transform the camera, but I'm not sure the Shogun is worth it on any mirrorless cameras. If the Shogun could record RAW off these cameras, then I could see a good reason to get it, but by these pixel peeping comparisons I don't think so.
Thomas Smet
Veteran
That haloing looks like what one would get from filtering or blurring the chroma channels. Very odd indeed. This would be murder for any type of VFX application where one was expecting to use the Shogun for keying shots.
Is the GH4 actually outputting 4:2:2 or is it filtered to make the output 4:2:2? I sure the heck hope it isn't filtered that aggressively.
Is the GH4 actually outputting 4:2:2 or is it filtered to make the output 4:2:2? I sure the heck hope it isn't filtered that aggressively.
i can't add shogun footage, but i did a litte test to estimate the impact of the shogun compression formats on raw gh4 4k data:
you can download the stills in full size and resolution here:
http://users.mur.at/ms/projects/banding-test/compare.zip
it's based on a series of 100 time-laps-shots. darktable and a minimum of noise reduction was used to extract a 16 bit PNG sequence. this images were used to produce video clips by ffmpeg in different formats. later i extracted the 25th frame of any clip for comparison.
the differences are very small. ProRes LQ looks a bit worse than all others.
it isn't an impressive demonstration, but it could hint, what we should expect from similar shots/codecs under ideal conditions (not much movement etc.)
you can download the stills in full size and resolution here:
http://users.mur.at/ms/projects/banding-test/compare.zip
it's based on a series of 100 time-laps-shots. darktable and a minimum of noise reduction was used to extract a 16 bit PNG sequence. this images were used to produce video clips by ffmpeg in different formats. later i extracted the 25th frame of any clip for comparison.
the differences are very small. ProRes LQ looks a bit worse than all others.
it isn't an impressive demonstration, but it could hint, what we should expect from similar shots/codecs under ideal conditions (not much movement etc.)
Last edited:
Schmeknicolor
Member
I'm a bit late but the problem here with the halowing, i think, is the "Exact same settings..." you're using, you should have more color information on 10bit so you should have to tinker the settings.
Thomas Smet
Veteran
i can't add shogun footage, but i did a litte test to estimate the impact of the shogun compression formats on raw gh4 4k data:
you can download the stills in full size and resolution here:
http://users.mur.at/ms/projects/banding-test/compare.zip
it's based on a series of 100 time-laps-shots. darktable and a minimum of noise reduction was used to extract a 16 bit PNG sequence. this images were used to produce video clips by ffmpeg in different formats. later i extracted the 25th frame of any clip for comparison.
the differences are very small. ProRes LQ looks a bit worse than all others.
it isn't an impressive demonstration, but it could hint, what we should expect from similar shots/codecs under ideal conditions (not much movement etc.)
Couple of things about your test:
1. Thank you for doing this. while I question its ability to really paint a real picture of what the quality difference actually is it is good to show these kind of tests so people realize that the in camera compression is not as bad as one would think it is. Compression doesn't always have to be a bad thing and I think in camera compression gets a bad reputation these days. A good compression format is typically designed to replicate what the camera is capable of as close as technologically and economically feasible. This of course is always open to debate since there are some that think they should get everything they want. The fact is however the compression methods done in camera are generally done with great care to optimize quality while balancing file size and realistic expectations for that camera.
At the end of the day external recording can clean up certain aspects of the image quality but it will never magically make a whole new camera or a brand new look. Even the 4:4:4 vs 4:2:2 vs 4:2:0 is splitting hairs these days when it comes to 4k. Those pixels are so tiny that it really is close to humanly impossible to notice 4:2:0 with the naked eye. The whole 4:2:0 has been largely blown out of proportion for years and is typically the immediate scapegoat for overall bad quality in keying operations. Of course 4:4:4 will always be better but the amount by which it is better is actually very small.
2. Now for the negative. While this can be a better representation of a scene with little to no movement it does not even come close to telling us what the codecs are doing under stress. IPB codecs work off of motion prediction. If nothing moves in the frame then the motion prediction is insanely easy to do. If there is a lot of random motion a IPB format such as the 100 mbps on the GH4 can start to show macro blocking. The only tru way to test compression formats is to do so with high levels of motion and complex detail. The easier the scene to compress the less one will notice between different formats. The more complex the detail and motion the more the Shogun will start to stand out. Shoot a bunch of surfers riding waves in the ocean and the Shogun will show its advantage and not have a picture that starts to fall apart.
It is really about consistency. Using a I frame only format allows one to know the compression quality will be the same no matter how complex the frame is. IPB formats can vary depending on complexity so there is a lack of consistency between shots. Now the higher the IPB bitrate the less you will notice this. If the GH4 used 200 mbps you would notice the difference much less. At 400 mbps you would be lucky to see any difference at all without using some extreme methods and tools to hunt for the differences.
3. Computer compression is not the same as camera compression. The share the same basic H264 concept but no two encoders are ever the same. Camera encoders tend to have to work very quickly and may not optimize quality as much as a computer encoder will. A computer doesn't have to worry about encoding 4k video in real time without any room for error. A camera has to make sure it can chug through the frames without any error. This is a much harder task and camera encoders tend to be slightly lower quality than what a computer encoder can do. It also depends on what computer software encoder you are using as each of those can have a different look. Even between two NLEs like Premiere and FCPX there is a quality difference between the H264 encoding.
4. Honestly the best way to know how Shogun footage will look at 4k is to look at the last ten years of external video recording. 4k, HD and SD has exactly the same benefit in terms of quality with resolution being the only difference. External recording is not some magical process and one can already see how much the footage will improve by looking at how HD and SD are improved (or not really improved) when externally recorded. The last camera I used that actually showed a visual advantage to external recording was the Sony F900 which recorded HDCAM internally. HDCAM was actually a pretty awful format compared to what we have today. 1440x1080 with 3:1:1 color (only slightly better than 4:1:1 color). At the time it was the best we had but I would sometimes tether a computer with a Blackmagic Multibridge during studio shoots in order to preserve the cameras 1920x1080 4:2:2 sensor. External recording really helped with HDV as well for many of the same reasons. This was also back when many of the projects I worked on were 1080i. 4:2:0 interlaced is very ugly and without a doubt external recording 4:2:2 is vastly superior to 4:2:0 interlaced. Ever since moving to full resolution encoding formats and true progressive 4:2:0 the advantage has greatly been reduced. H264 also happens to be at least 2x better at compressing video compared to mpeg2. 50 mbps HD on my GH3 really does look as clean as 100 mbps HD mpeg2 if such a camera format existed. Considering HDV was 24 mbps and mpeg2 one can see how much less we need external recording today in order to maintain optimal quality or at least pretty darn close to optimal quality.
thanks for such an extensive and profund reply!
in agree with you an nearly every aspect -- even the more critcal/negative ones!
you articulated a lot of very important details, i could never express so clear.
i published this pictures here, because is was not very satisfied with neumanns comparison shots. shure, i really liked this first available objective output from both sources, but the also triggered a lot of new questions in my mind. i could not explain to myself, why this material doesn't look better? i did my time-laps experiment a few days earlier and i was expecting better results -- in particular, recording static scenes like in my experiment.
when i did my shots, i was chiefly interested in the impact of 8bit vs. 10bit image representation and it's impact on banding issues. i used the time-laps detour instead of one continous frame, to have some realistic noise in the resultant clips. sensor noise and grain are very important when it comes to the perception of banding defects. and they also should be seen as relativizing measurement with regard to the importance of higher dynamic representation. shure -- you can not get this aspect from this pictures, i just reused them, but i'm more deliberating abbout this other question until now.
so, i was well aware that my test are not very representative, but i wanted to stimulate the discoussion, because this prtical output looked so disappointing to me.
yes -- i perfectly agree!
and i think, it is very important to emphasize the fact, that actual cameras to this job incredible well by using dedicated hardware support. i'm iin doubt about seeing similar H.264 realtime processing from platforms like the shogun or odyssey without special hardware implementations. prores is quit simple in this respect. it may work very well in everday work when using the higher quality profiles and a lots of disk space, but from a technical point of view it's quite primitive compared with more actual codec development.
if diskspace and i/o-load isn't taken into account prores will work very well -- like jpeg is still a very convenient image compression format even though we have more sophisticated technics now. prores has lots of benefits in it's primary rolle as intermediate format, but this shouldn't be tacitly extented to other utilisations. if you want want to archive your original footage, size matters a lot.
if we realy want to get out of the camera what it could, i/o-bandwith and codecs of reasonable bandwith are the most limiting factors.
i don't think, relative cheap and simple external hdmi-recorders should be seen as the one and only way to solve this problem. why shouldn't cameras utilize very common actual consumer computer technology to store their footage themself via USB3 on external SSD? this would give us min. 300-400Mbps recording capability on devices of aedequate size without all this complicated and error prone roundtrip over hdmi. and it would only need simple cheap consumer technology, not expensive very reliable and state of the art professional storage equipment. it should not be impossible to record that way as good as the shogune or even better without reduplicate everything (including sources of trouble) on two devices.
ok -- this is wishfull thinking, but it would make more sense to me than this very limited actual improvement.
you are perfectly right! -- i should have add H.264/H.265 all-intra examples too, even if we can not use them on the GH4 in 4k mode right now.
but this more recent codec do not use only motion prediction, they also utilze a lot of intra frame prediction!
that's the main advancement over prores et.al. (look at http://bellard.org/bpg/ how impressive this works!)
but as a matter of fact all this very advanced new compression technologies are most suitable for very high compression ratios. if you are more interested in very high quality results the computing requirements gets challenging and the benefit in size and visible quality isn't that impressive anymore.
but if you want to optimize size and quality at same time, it's the way to go.
professional implemations like XAVC or AVC-Ultra 4K are very well actual examples.
yes -- i have to agree again!
but it's very interesting to study and recognize this differences beween optimal implementations and the real output from existing soft-/hardware.
that's why i'm indeed interested in this shogun/GH4-internal comparison: it opens a entirely new sight inside the black box of GH4s internal image processsing. this can be very useful to optimize workflows and use of this camera even for non-shogun-owners!
in agree with you an nearly every aspect -- even the more critcal/negative ones!
you articulated a lot of very important details, i could never express so clear.
i published this pictures here, because is was not very satisfied with neumanns comparison shots. shure, i really liked this first available objective output from both sources, but the also triggered a lot of new questions in my mind. i could not explain to myself, why this material doesn't look better? i did my time-laps experiment a few days earlier and i was expecting better results -- in particular, recording static scenes like in my experiment.
when i did my shots, i was chiefly interested in the impact of 8bit vs. 10bit image representation and it's impact on banding issues. i used the time-laps detour instead of one continous frame, to have some realistic noise in the resultant clips. sensor noise and grain are very important when it comes to the perception of banding defects. and they also should be seen as relativizing measurement with regard to the importance of higher dynamic representation. shure -- you can not get this aspect from this pictures, i just reused them, but i'm more deliberating abbout this other question until now.
so, i was well aware that my test are not very representative, but i wanted to stimulate the discoussion, because this prtical output looked so disappointing to me.
yes -- i perfectly agree!
and i think, it is very important to emphasize the fact, that actual cameras to this job incredible well by using dedicated hardware support. i'm iin doubt about seeing similar H.264 realtime processing from platforms like the shogun or odyssey without special hardware implementations. prores is quit simple in this respect. it may work very well in everday work when using the higher quality profiles and a lots of disk space, but from a technical point of view it's quite primitive compared with more actual codec development.
if diskspace and i/o-load isn't taken into account prores will work very well -- like jpeg is still a very convenient image compression format even though we have more sophisticated technics now. prores has lots of benefits in it's primary rolle as intermediate format, but this shouldn't be tacitly extented to other utilisations. if you want want to archive your original footage, size matters a lot.
if we realy want to get out of the camera what it could, i/o-bandwith and codecs of reasonable bandwith are the most limiting factors.
i don't think, relative cheap and simple external hdmi-recorders should be seen as the one and only way to solve this problem. why shouldn't cameras utilize very common actual consumer computer technology to store their footage themself via USB3 on external SSD? this would give us min. 300-400Mbps recording capability on devices of aedequate size without all this complicated and error prone roundtrip over hdmi. and it would only need simple cheap consumer technology, not expensive very reliable and state of the art professional storage equipment. it should not be impossible to record that way as good as the shogune or even better without reduplicate everything (including sources of trouble) on two devices.
ok -- this is wishfull thinking, but it would make more sense to me than this very limited actual improvement.
you are perfectly right! -- i should have add H.264/H.265 all-intra examples too, even if we can not use them on the GH4 in 4k mode right now.
but this more recent codec do not use only motion prediction, they also utilze a lot of intra frame prediction!
that's the main advancement over prores et.al. (look at http://bellard.org/bpg/ how impressive this works!)
but as a matter of fact all this very advanced new compression technologies are most suitable for very high compression ratios. if you are more interested in very high quality results the computing requirements gets challenging and the benefit in size and visible quality isn't that impressive anymore.
but if you want to optimize size and quality at same time, it's the way to go.
professional implemations like XAVC or AVC-Ultra 4K are very well actual examples.
yes -- i have to agree again!
but it's very interesting to study and recognize this differences beween optimal implementations and the real output from existing soft-/hardware.
that's why i'm indeed interested in this shogun/GH4-internal comparison: it opens a entirely new sight inside the black box of GH4s internal image processsing. this can be very useful to optimize workflows and use of this camera even for non-shogun-owners!
Tom Antos just posted these, ungraded and heavily graded. It's not night and day but I do see a bit of difference between them on the graded one.
https://vimeo.com/116275533
https://vimeo.com/116275534
https://vimeo.com/116275533
https://vimeo.com/116275534