Otis13

Joined on Dec 18, 2020

Comments

Total: 36, showing: 21 – 36
« First‹ Previous12Next ›Last »
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Internet Enzyme: So the main point that I'm seeing this article make, is that the improvement delta between gamma encoded rec. 709 video and raw video is not analogous to the improvement seen in photography when we move from JPEG to raw because of the fact that JPEG is a comparatively even worse acquisition format, which is more of an academic argument than a pragmatic point, I think. The second point is this kind of tired, typical set of technical and expense considerations that people love to raise regarding this topic, but I feel that these should not discourage anyone who is actually aiming to achieve the highest quality image they can: if you have the budget, there's no reason you shouldn't be shooting raw. Then there are the issues with Final Cut and ProRes RAW which I feel are too specific. Perhaps the scope of this should have been limited to ProRes RAW and Final Cut specifically, but I suppose that limits the potential clicks

Mapping is not compression. You can have a container in a given precision, and convert between these LUTs in the very same container. You can go from high precision to low precision and the opposite. It's not about fewer of more. There are precision problems when changing containers and doing mapping, but this happens in any representation.

In my original post I wanted to clear the discussion about these three categories. They represent formats and workflows, and not compression or precision. These mappings can be used with any compression and precision. I was talking about substance and I know how imaging works. You clearly do not know how imaging works and you were discussing semantics.

Link | Posted on Feb 17, 2021 at 01:52 UTC
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Internet Enzyme: So the main point that I'm seeing this article make, is that the improvement delta between gamma encoded rec. 709 video and raw video is not analogous to the improvement seen in photography when we move from JPEG to raw because of the fact that JPEG is a comparatively even worse acquisition format, which is more of an academic argument than a pragmatic point, I think. The second point is this kind of tired, typical set of technical and expense considerations that people love to raise regarding this topic, but I feel that these should not discourage anyone who is actually aiming to achieve the highest quality image they can: if you have the budget, there's no reason you shouldn't be shooting raw. Then there are the issues with Final Cut and ProRes RAW which I feel are too specific. Perhaps the scope of this should have been limited to ProRes RAW and Final Cut specifically, but I suppose that limits the potential clicks

You seam to care a lot about semantics. A format is the way something is arranged. Log format is a common combination of words. Google it. Visit the Fujifilm website, and you will find "output to the HDMI device in F-Log format."

I use LOG format in general discussion, LUT (programmers), and gamma (retired videographers). All these things are a simple look up table that gets a precalculated value from a set of values stored in memory. A y=f(x) for a x range.

It's not a tonal compression, it's a mapping. How all the the different formats came along is a very long story, but simply put, motion picture film DI LOG workflow, linear RAW still photograph workflow, and video recording merged in our times. The CRT display gamma is required for getting back to linear light that looks like the scene, what the sensor actually captures and it's a lot more complex than this. Everything else is a mapping.

Link | Posted on Feb 16, 2021 at 23:07 UTC
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Internet Enzyme: So the main point that I'm seeing this article make, is that the improvement delta between gamma encoded rec. 709 video and raw video is not analogous to the improvement seen in photography when we move from JPEG to raw because of the fact that JPEG is a comparatively even worse acquisition format, which is more of an academic argument than a pragmatic point, I think. The second point is this kind of tired, typical set of technical and expense considerations that people love to raise regarding this topic, but I feel that these should not discourage anyone who is actually aiming to achieve the highest quality image they can: if you have the budget, there's no reason you shouldn't be shooting raw. Then there are the issues with Final Cut and ProRes RAW which I feel are too specific. Perhaps the scope of this should have been limited to ProRes RAW and Final Cut specifically, but I suppose that limits the potential clicks

I have found that most people are more comfortable with LOG and RAW processing, so one of the two formats will prevail as a standard. Manipulation in the display LUT is a rare art these days and it depends on the camera specifics, because all cameras are different. A manufacturer provided conversion in post to get from display LUT to LOG and Linear would get better results than "RAW" and log at the same datarate.

Link | Posted on Feb 16, 2021 at 21:42 UTC
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Internet Enzyme: So the main point that I'm seeing this article make, is that the improvement delta between gamma encoded rec. 709 video and raw video is not analogous to the improvement seen in photography when we move from JPEG to raw because of the fact that JPEG is a comparatively even worse acquisition format, which is more of an academic argument than a pragmatic point, I think. The second point is this kind of tired, typical set of technical and expense considerations that people love to raise regarding this topic, but I feel that these should not discourage anyone who is actually aiming to achieve the highest quality image they can: if you have the budget, there's no reason you shouldn't be shooting raw. Then there are the issues with Final Cut and ProRes RAW which I feel are too specific. Perhaps the scope of this should have been limited to ProRes RAW and Final Cut specifically, but I suppose that limits the potential clicks

It would be better if we called these formats "Linear", "Log", "BT.709 or display". That's their real difference. The linear format typically leaves white balance and demosaicking for post, but that's not really an advantage in the present workflows.

Link | Posted on Feb 16, 2021 at 20:51 UTC
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Vladyslav Kosulin: The main confusion is that all these 'RAW' video codecs are not really RAW. These are codecs that pretend to preserve as much data as possible while making bitrates and file sizes manageable.
As far as I know ARRI is the only camera maker that still offers true RAW capture format for 4K, but no camera records true RAW for 5.5K and up. Everybody downconverts to 10-12 bit with or without log, and with various lossy compression options.

By full resolution I mean continuous pixels. Cropping is still full resolution at the cropped area. Skipping pixels presents many mathematical problems.

I use an XT4 is 400mbit ISO"640" DR400 and full shadow tone boost. This is the minimum ISO really, an ISO160 with about 2 stops boost and shaping in the highlights to preserve the full ISO160 range. The in-camera recording format gets everything the sensor has to give, all the highlights (of course) but also all the shadow and the full noise of the sensor. The PSNR is high enough to do corrections in post, and the compression precision is good enough for applying heavy processing. It really is as good as RAW. There is only a disadvantage in underexposure, but underexposure is a waste of dynamic range in RAW as in any other format, RAW merely reduces compression artifacts in the underexposed image, and only if it's uncompressed RAW.

Link | Posted on Feb 16, 2021 at 20:24 UTC
On article Sony a7S III review (2782 comments in total)
In reply to:

Otis13: DPREVIEW, it's time you add a full ISO series for 4K VIDEO in all reviews. Video is a main feature now.

Sony have power zoom lenses for E-mount, but this is again a videography feature. For Fuji/Sony there are manual lenses that are designed for cinematography and are very affordable. There are many primes and there is the Fujinon MK/MKX zoom line that is designed for manual control and quite affordable.

Link | Posted on Feb 16, 2021 at 20:08 UTC

Hare Ball wins in every single category. Not even competing with the others.

Link | Posted on Feb 16, 2021 at 09:50 UTC as 6th comment | 2 replies
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

Vladyslav Kosulin: The main confusion is that all these 'RAW' video codecs are not really RAW. These are codecs that pretend to preserve as much data as possible while making bitrates and file sizes manageable.
As far as I know ARRI is the only camera maker that still offers true RAW capture format for 4K, but no camera records true RAW for 5.5K and up. Everybody downconverts to 10-12 bit with or without log, and with various lossy compression options.

There is only one true RAW. Linear light, full bit depth, full pixel resolution, uncompressed data from the sensor, because this is what the sensor outputs, and this is the only way to postpone the processing decisions. That is about 300MB/sec for 4K 12bit 24p. A modern SSD can record and playback one track.

It's actually an excellent format because it's easily processed in realtime to 1080pRGB for display and editing, but not many people want to deal with, due to the storage requirements and the RAW processing quality which is grossly inferior to in-camera algorithms coming from the big manufacturers.

I supported and used RAW when it was extremely rare. Today it's just a marketing thing. The same applies to log formats. If it will be compressed, a LUT that matches the display LUT with some extra dynamic range shaped in the highlights is technically ideal. This can be converted to linear if a RAW workflow is required, or LOG. It will be better than "RAW" or LOG at the same datarate.

Link | Posted on Feb 16, 2021 at 00:21 UTC
On article Why Raw video might not be the game-changer you expect (257 comments in total)
In reply to:

MikeRan: Thanks for clearing up so many misconceptions. Never really understood the draw to RAW over a good, gradeable profile with 10-bit encoding.

In reality it's worse than this. Even in base ISO, the color matrix eats up some of the dynamic range, up to two stops in very warm lighting for blue. In higher than base ISO it's every worse. Technically speaking, a full 8bit of RGB in a LUT appropriate for display is something very few cameras can achieve in their base ISO and only in daylight balance. That covers about 11 stops of linear dynamic range in a modern LUT.

In modes that allow extra room in highlight, the situation is even worse, because the noise floor is raised.

The dynamic range claims by manufacturers are nonsense. Ideal case scenarios in multiple ways and technically inaccurate.

Meanwhile, the PSNR of in-camera recording formats is improving. That means they actually have good imaging quality and perfectly usable shadows. Some in-camera formats today have no blocking if the shadows are boosted a little bit in camera.

The RAW approach itself is inferior in multiple ways. Inferior demosaicking and processing.

Link | Posted on Feb 16, 2021 at 00:00 UTC
On article Sony a7S III review (2782 comments in total)
In reply to:

Otis13: DPREVIEW, it's time you add a full ISO series for 4K VIDEO in all reviews. Video is a main feature now.

Many cameras of this class are perfectly capable of shooting feature films, art projects and music videos, and more practical that any cinematography solution of the past. By a proper video camera you probably mean a camera for event videography, run and gun, etc. These are things I'm not interested in.

Link | Posted on Feb 15, 2021 at 21:35 UTC
On article Sony a1 review (2564 comments in total)
In reply to:

deepDEEPpurple: And just like that, Sony made Canons effort look redundant. Canon can't catch a break.

Few months ago R5/8k were going to revolutionize the world.

Quickly cracks started to appear: overheating, banding in shadows, terrible rolling shutter, artificial limitations and performance degradation due to heat.

The hype came crashing down: "8k is too difficult", "no one can make a camera that shoots 8k and is good as stills".

Sony went: Hold my beer!

One by one, all R5 problems were fixed by Sony:
-GeraldUndone called the R5 "unusable" and the A1 "better than A7rIV and better video camera than A7sIII. R5 is worse than all of them in their respective fields"
-iPhonedo recorded 1h30m of 8k
-BIF don't won't look warped anymore due to RS
-No artificial video limitations (you can film forever with ext. power as done by GU)

Sony leaped all Canon offerings effortlessly. Why buy a R5, C70 and 1Dx when the A1 will do everything

Only better, Faster and 0 overheat

Now, who wants to buy a kidney?

That's not how it works. The color of a camera depends on the original color which is mapped to get what you see in images. The spectrum of each of the R, G, B filters in the pixels of these cameras defines what the image will look like. This cannot be changed by a matrix. A matrix merely subtracts a part of the other two colors to get to a type of primaries. No raw processor can truly match color. That's why most cameras have very bad reds. Almost all of them are too orange. Have you noticed Sky cyans. you rarely get a good shade. You are not doing your own color, you start with what you get from the sensor primaries, as processed through a matrix. If what you believe was true, all cameras would have flawless color. They can barely get some skin tones right and allow everything else to deviate.

Link | Posted on Feb 15, 2021 at 03:06 UTC
On article Sony a7S III review (2782 comments in total)
In reply to:

Otis13: DPREVIEW, it's time you add a full ISO series for 4K VIDEO in all reviews. Video is a main feature now.

When we compare video quality in imaging, we use PSNR in a series of frames but even a compressed still is very useful in intraframe formats.

The main problem will be interframe compression. On a static scene, stills provide very useful information, but in some cameras the first GOP range is faulty.

Link | Posted on Feb 14, 2021 at 22:54 UTC
On article Sony a7S III review (2782 comments in total)

DPREVIEW, it's time you add a full ISO series for 4K VIDEO in all reviews. Video is a main feature now.

Link | Posted on Feb 6, 2021 at 20:12 UTC as 64th comment | 8 replies

115mmx64mm (in 16:9), 2922x1644 resolution (in 16:9). 5.2" diagonal.

Link | Posted on Jan 27, 2021 at 09:10 UTC as 31st comment
On article Fujifilm XF 10-24mm F4 R OIS WR updated sample gallery (33 comments in total)
In reply to:

(unknown member): It's quite nice on the wide end but the closer you get to 24mm the poorer the IQ. I don't think people generally buy this lens to shoot at 24mm so maybe that's not a deal breaker but at this price an f4 lens should be sharper through the entire range IMO.

Personally, I find myself shooting my 16-35 FF lens at 35mm fairly often so I think I'd have to pass.

It's still a nice option for video/vloging if that's your main interest.

Fuji publish MTF for their lenses. One can easily find out that 10-24 at telephoto has very low MTF at the corners.
https://2.img-dpreview.com/files/p/TS560x560~forums/52737770/b2da7c6d26fd452ebd91456d7ec11ee5

Link | Posted on Jan 27, 2021 at 08:26 UTC
On article Fujifilm XF 10-24mm F4 R OIS WR updated sample gallery (33 comments in total)
In reply to:

(unknown member): Appreciate the effort by DpReview to redo the testing with a new sample.
The image quality is improved.
We still have to remember that @10mm the corners will be soft for an ultra-wide lens but otherwise it is an interesting lens that gives us another perspective and now with WR.

It depends on the design. The edges are soft in simple designs using less elements. 10-24 is still very good for the number of elements it has and for its cost. The 8-16 is much better but it has 20 elements, not 14.

Link | Posted on Jan 27, 2021 at 07:59 UTC
Total: 36, showing: 21 – 36
« First‹ Previous12Next ›Last »