Bern Caughey
Veteran
Will 400Mbps be available for 1080p? Only heard those data rates associated with 4K.
GH4 'Lab Review Reveals 10bit Flaws in Vlog + FREE GH5 LUT'
- Thread starter Bern Caughey
- Start date
Will 400Mbps be available for 1080p? Only heard those data rates associated with 4K.
Thomas Smet
Veteran
It is different. Don't just take my word for it. Look at the cinema cameras that shoot raw and ProRes. Even the Arri Alexa can record ProRes and some Hollywood films have used it instead of shooting raw to make the post work go faster.
Just saying both are 800mbps and both use DCT is way over simplifying what codecs do to images. There is no magic under the hood with ProRes. In fact it is the opposite. The magic is in H264 because it does a lot of stuff to help reduce the files. Besides even the same DCT format like Jpeg can end up with different results depending on which software created it. DCT is the process and not a guarantee of how something will look. Its like saying farmer A and farmer B will produce the same amount during a harvest because both are farmers. Well yes in general they should produce about the same but there are other factors that play into their harvest like skill, dedication, equipment, soil and so forth.
Don't believe me then shoot some 200mbps all-I HD video on the GH4 and then compare that to external captured ProRes. ProRes422 HQ recording HD at 30p uses 220mbps so really about the same yet the quality difference is pretty significant. The proof is right there that all-I h264 even at the same bitrate doesn't hold up as well to ProRes. In fact ProRes422 LT at only 102 mbps looks better. I'm not talking about the difference between 10bit and 4:2:2 either but the overall quality of the image.
The same is true of Mpeg2 all-I which is 4:2:2 but only 8bit. That format can go up to 300mbps and in theory should blow away ProRes422 HQ at only 220mbps but it does not.
Plenty of examples and tests have been done for over a decade on the benefits of a format like ProRes over mpeg2 and H264 all-I formats. Take AVC-Intra and AVC-Ultra for example. Darn impressive formats but still not visually as good as ProRes. I can't tell you exactly what Apple does to make their codec so robust but the same is true for Avid DnxHD. People have even compared those two formats to each other and noticed differences. On paper there shouldn't be any difference between DnxHD and ProRes yet there is.
Thomas Smet
Veteran
Have you ever looked at any GH4 V-log material recorded externally? Plenty of people have and can see just how much damage the H264 does. You can send 8bit V-log through the HDMI to an external recorder as well and it looks vastly better than the internal 8bit.
Cary Knoop
Veteran
Last edited:
Cary Knoop
Veteran
Seriously now, how can that possibly be an argument? Did I say ProRes is bad?
Of course the output is great, it's a high end cinema camera after all, or would you claim a big part of it is due to ProRes?
I am sure if it could render H.264 with a high bitrate it would look good as well!
Sorry but this argument to me is as valid as saying use an Arri LUT for your GH5 because, well, ARRI uses it, so it must be good.
joe12south
Veteran
No, the noise (and detail) is not being smushed away by h264.
Thomas Smet
Veteran
You fail to accept the points I'm trying to make based on a decade of tests, studies and examples. You want to believe H264 can be just as good but where is your proof? I gave you plenty of examples of all-I mpeg2 and what we have experienced with the GH4 and you seem to not want to accept that at all. If you want to argue then fine provide some proof that all-I H264 can look exactly the same as ProRes. I'm waiting.
Do you deny that the 200mbps all-I HD format on the GH4 doesn't look as good as ProRes HD? We don't even have to wait for the GH5 to see an example staring us in the face that proves my point. Why are you not accepting these clear examples I am providing? Almost every Gh4 shooter knows the 200mbps format on the Gh4 isn't worth using. That is the same bitrate as ProRes HQ HD so it should look amazing but it does not.
I would love to be proved wrong but after a decade of studying image processing I have yet to see a format like mpeg2 or H264 all-I match the visual quality of ProRes. In fact other similar formats like photoJpeg which according to you should be exactly the same because they are DCT based should be identical but yet they are not.
My point about cinema cameras is not that ProRes makes them great but that ProRes is used because it does a better job of maintaining the quality the sensor and image processing can produce in those cameras. Most film makers wouldn't be caught dead using an H264 format no matter how fancy one tries to make it. ProRes was selected because of industry support and because it holds up better than many other formats. Again I don't know why exactly ProRes holds up better but it does. I am basing this on a decade of hundreds of people studying the benefits of ProRes as a cinema format.
Here are two examples. One recorded ProRes HQ and the other converted to the Panasonic professional AVC-Intra 100 format. This is a 4:2:2 10bit all-I H264 based professional format. Look at the extra mosquito noise and blocking in the AVC-Intra example. This was a GH4 V-log shot with the Leeming V-log lut applied. I am showing an area of flat color because this is ironically where H264 falls apart because it assumes the colors are too close together because of log. ProRes doesn't even think about flat areas of color. It just looks at the pixels and treats every one equally.
Cary Knoop
Veteran
If your point is that ProRes is a good codec you are making a strawman argument, I never claimed that ProRes is not a good codec.
If your point is that ProRes is better because if you compare it with H.264 with a far lower bitrate then you are making a false comparison.
When you make comparisons between codecs you should compare them with similar bitrates and inter/intra frame options.
Everybody understands that comparing one codec with a bitrate of X with another codec with a bitrate twice or even more than twice is not a fair comparison.
I think we should at this point simply agree to disagree.
morgan_moore
Major Contributor
It is worth considering why h264 has a history of being 'bad' .. it is doubtless a great codec typically configured with huge compression
My EX1 had 2 8gb cards.. god knows what they cost, now if you can buy 8gb cards at all it is from the Walmart bargain bin.
264 has a history of being used where the manufacs assumed the user needed to skim data to the minimum at every possible juncture. To some extend the 'facs are behind users found on this board in that they are still trying to deliver data load/image quality in a ratio that we don't want. BMC deliver the 'indy' data heavy cameras for art boy shortform.
Ultimately we pay our money and take our choice. Of course a few tweets at a 'fac asking for more data cant do damage.
My EX1 had 2 8gb cards.. god knows what they cost, now if you can buy 8gb cards at all it is from the Walmart bargain bin.
264 has a history of being used where the manufacs assumed the user needed to skim data to the minimum at every possible juncture. To some extend the 'facs are behind users found on this board in that they are still trying to deliver data load/image quality in a ratio that we don't want. BMC deliver the 'indy' data heavy cameras for art boy shortform.
Ultimately we pay our money and take our choice. Of course a few tweets at a 'fac asking for more data cant do damage.
Thomas Smet
Veteran
Dude I have made comparisons at similar bitrates. Have you actually been reading anything I posted?
GH4 200mbps HD all-I. ProRes HD is 220mbps at 30p. On the GH4 it looks horrible.
Mpeg2 300mbps HD all-I. Should be much better or at least equal to ProRes HQ. It is not.
ACD-Intra is lower at 100mbps but it is a professional grade codec that is 10bit and used by a lot of broadcasters world wide. Besides ProRes LT HD uses 102mbps for 30p as well so it should be compatible to that but it isn't.
And what exactly has been this point the entire time. That eventually the GH5 will get 400mbps all-I which is comparable to AVC-Intra HD at 100mbps. Showing AVC-Intra 100 is the real world comparison to the best the 400mbps all-I could possible look on the GH5. Even if Panasonic did cram AVC-Ultra (the 4k version at 400mbps) into the Gh5 (which I highly doubt they did) it still wouldn't look as good as ProRes which we can see in the examples I provided.
If you have any evidence otherwise I would love to see it. I love to be proven wrong. Otherwise you are just blowing smoke.
I think Thomas is pointing out that they are at the same bit rate. Also all H264 encoders are not the same. Try a few yourself they are all very different the best viewed to be X264.
hyalinejim
Well-known member
https://www.youtube.com/watch?v=AtVgFj_G-q4
- It's undecided yet whether the anamorphic mode will have the option to desqueeze or not
- The "6K" 24p anamorphic update due in the summer will use the full width and height of the sensor at a resolution of approx. 5000 x 4000 in h265 (although not 400Mbps h265, it will be a lower bitrate)
So, assuming that the h265 implementation is better than 150Mbps h264 IPB and/or 400mbps h264 All-I, this could be a great option for those who want to shoot V-Log internally without chroma smearing artifacts... as long as you're willing to shoot 4:3 and crop top and bottom when using a non-anamorphic lens.
- It's undecided yet whether the anamorphic mode will have the option to desqueeze or not
- The "6K" 24p anamorphic update due in the summer will use the full width and height of the sensor at a resolution of approx. 5000 x 4000 in h265 (although not 400Mbps h265, it will be a lower bitrate)
So, assuming that the h265 implementation is better than 150Mbps h264 IPB and/or 400mbps h264 All-I, this could be a great option for those who want to shoot V-Log internally without chroma smearing artifacts... as long as you're willing to shoot 4:3 and crop top and bottom when using a non-anamorphic lens.
Agreed. Which is why I emphasized "could" when I say that 400 Mb/s could make a worthwhile difference here. I've no illusions that it will, only that it's...possible.
Last edited:
I actually did not find it so, and even less so in V-Log. Which is part of the reason I was hoping for a 400 Mb/s All-I mode on the GH5.
I'm glad we got it, and eagerly await the nitpicky testing in September.
Those are helpful, but often I find it more helpful to see examples of where it's not so good. I find it useful to see tests and examples that push camera and codec to the breaking point. This helps me decide if the flaws (and all cameras have them) are something I can live with, or if I need to keep looking. The flaws in the GH4's internal V-Log forced me to keep looking.
So I'm interested if the flaws in the 400 Mb/s V-Log will do the same, or will I once again be a Panny-fan.
Last edited:
Thomas Smet
Veteran
Why would you assume the h265 implementation is better? Its a very new codec and very processor intensive. When h264 first appeared in cameras it wasn't as good as it would eventually become after a few generations.
Compression is compression when dealing with very similar colors like gradients in skin or the sky. h265 is a giant step in the opposite direction by compression material even more than it is with h264. It is efficient and impressive by what it can do at lower bitrates but tossing out information will always, always always come at a cost. We may not see it with the naked eye but trust me it is there. Any log format creates an image with may more subtle changes in color than what an encoder was ever really meant to handle. An encoder is really optimized for a real world looking image and not the stretched subtle faint color and contrast changes of a log format.
Honestly log is better suited to raw and uncompressed (I mean real uncompressed, not lightly compressed formats like ProRes and DnxHD). At the very least ProRes422 HQ (not the other flavors) can do a really good job but depending on the detail of the shot still may not be perfect. Every step we take in the opposite direction to compress video more and make the sizes smaller will introduce a whole new set of odd subtle loss.
Thomas Smet
Veteran
Kind of depends really. Any camera can be made to look bad. That isn't an indication of what a camera is capable of. If some samples are great that is more of a true representation of how good it can be. If you want to aim for the lowest common denominator and judge a camera used in the worse possible way thats fine but I tend to look for what a camera is capable of and not how bad it can look in the wrong hands. There are people that can make an Arri Alexa look like garbage and it would be silly to judge it based on those situations.
The funny thing about cameras, graphics software, 3D animation tools and so forth is that the professional tools tend to get used by the seasoned pros so we see a much higher percentage of amazing material from the higher end tools. Lower end tools like open source design tools or cheaper cameras are used by a mixture of talented professionals and untrained users so we get a mixture of results across those tools. Doesn't mean they are bad. Just that our perception of them and the examples we see may not always be the same caliber. That is why I always look towards what a camera is really capable of at its absolute best. That is the higher benchmark I want to aim for and not the lower benchmark .
Cary Knoop
Veteran
Raw video is not log.
You store log video uncompressed? Which with camera and recorder?
Honestly I think it is complete overkill to store uncompressed video and there is zero, not even a theoretical, advantage over using lossless compression.
Last edited:
Both H.264 and H.265 were optimized for commercial delivery of highly compressed HD and 4K video products. They perform best when used to compress professionally graded Rec 709 content (or Rec 2020 for H.265) using two or more passes to analyze abrupt scene transitions in advance. Used properly, they will maintain consistent image quality within constrained bitrates suitable for streaming as well as file-based storage.
Neither of these codecs are particularly well-suited for real-time video image capture. Efficient H.26x compression relies on intensive geometric searches for motion-tracked intra and inter-frame image patterns and deletion of imperceptibly fine details. In a video camera, this must be done using fixed internal frame buffers within strict bitrate and timing restrictions dictated by the frame rate. To insure reliable image capture, a real-time encoder must cautiously use only the most predictable compression techniques rather than employ more sophisticated techniques that require large memory buffers and extended analysis time. As a result, much image detail is lost to suboptimal utilization of the encoder's hardware-constrained bitrate.
The reason Panasonic chose to use H.264 was because the GH line is a prosumer product intended to produce consumer videos directly viewable on HDTV's without further post-production. V-log L and 10-bit 4:2:2 color depth were not anticipated in the original design, they were enhancements retrofitted into video capture formats that were already marginally inadequate for their original purposes. They will work fine for consumer-grade purposes, but their technical shortcomings will become visible when closely examined using high quality post-production tools. For optimal image quality, use of a high-bitrate external video recorder is essential.
Last edited:
walkaboutcamera
Active member
All blackmagic non studio cameras when recording raw, get 16 bits from the sensor, then log encode the data to 12 bits. the bmpcc then writes lossless compreessed , the irsa s give you the choice of lossless compressed or lossy compressed.