Attila Bakos

I was experimenting with a few anti-aliasing scripts in avisynth to see if I could clean up moiré & aliasing in X-H1 120p footage. It's not perfect but it makes the footage more usable. Download a comparison from here: DOWNLOAD

On the left you see footage recorded in 120p F-Log (provided by Fenchel & Janisch), with my Classic Chrome LUT + a bit of contrast applied.
On the right you see the same color correction with additional anti-aliasing, temporal noise reduction and post sharpening.

Check the top of the bridge, less aliasing there. Also, the background buildings show less moiré. With more optimization the results can be even better.
The only downside of using avisynth is, well, avisynth itself. It's geeky stuff, not suited for everyone.

17 minutes ago, Prandi said:

If I film for so long then I hang it on the net ... then I need no battery at all :-)

You can loose short takes as well, if you forget that you started recording with low battery level on the currently used battery.

19 minutes ago, Andrew Reid said:

I'm debating whether to get the grip. Does it run off 3 batteries in total with the grip attached? So one is in the camera, and two in the grip. I believe battery life to be quite a bit above average (when not in boost mode), for a mirrorless camera. Although for stills, vs a Nikon D5, obviously it isn't going to run as long on a single battery. 900+ shots on the grip though, I believe?

Keep in mind that it can't switch batteries when recording, which means recording will stop when a battery runs dry, so you have to plan accordingly.

@BTM_Pix I noticed this in many X-H1 IBIS videos, I believe the only solution is to hold it as steady as you can (ninja walking, no running) or use an additional gimbal, otherwise you will have these sudden corrections when the IBIS has no more room to hold the image in a fixed position. It's still better than IBIS on full frame.

17 minutes ago, mkabi said:

I could be wrong here. But the first I heard of Dual ISO is back in 2013 and it was implemented by Magic Lantern, a year later Varicam comes out. Now I'm not saying that they copied ML... but really... what is Dual ISO??? It really is a photography technique -> Bracketing.

If anything I blame RED for copying ML???

And if we are playing the blame game.... pretty much everyone is copying RED when they implement 4K. Because, when everyone was saying 1080p is the end game - RED was pushing the envelop saying 4K son! And, that was back in 2006.

That dual ISO mode ML implemented is not the same as having two native ISOs. The first one is basicly a kind of bracketing, the second one is two different circuits dedicated to a lower and a higher ISO, before gain is applied. These two native ISOs have about the same noise levels, which is obviously not the case with ML's dual ISO.

41 minutes ago, frontfocus said:

I can do some more testing. I'd have to probably get a better subject and use a tripod to really compare it, but looking at the video you linked, to me it looks like he messed up his settings or files and only looked at slow mo from the X-H1, which seems to be pixel binned and not downscaled

Thanks, that would be much appreciated!

@frontfocus I had the same feeling so asked the guy specificly about the high speed setting, and he said it was switched off at the first shot.

@frontfocus Thanks for the test! I'm happy for the results but it's also confusing, that video earlier showed a massive difference in 1080p60 mode. I just checked it again, the movement is about the same, but the X-T2 looks noticably cleaner. There must be a difference in settings, or maybe his unit is flawed, I don't know. I don't really care about 120p, but as I'm switching from the X-T2, I want to be sure that I get at least the same 60p quality. Too bad I can't rent this camera where I live.

On 2018. 03. 21. at 7:00 PM, Trek of Joy said:

I believe the only solution is what's in the previously posted video - change the date to force the pixel remapping in order to do it manually. To be fair, its only happened once on each body I've owned. As to the a7III, I can't speak to anything other than the way it looks on videos I've seen since the camera isn't available in the US yet.

chris

The problem surfaced again in the newest video of the guy who posted the fix, and he told me the fix was unfortunately temporary, he's thinking about sending his A7III back.

On 2018. 03. 14. at 1:59 PM, Attila Bakos said:

@frontfocus Can you please check if 1080p60 has more moiré/aliasing on the X-H1 than on the X-T2? I'd like to know if you can confirm this.

On 2018. 03. 14. at 2:02 PM, frontfocus said:

@Attila Bakos I will do. Don‘t know if I find enough spare time today or tomorrow, since I first have to get my X-T2, but weekend at the lastest I will do

@frontfocus Hi, did you have time to check this?

38 minutes ago, BTM_Pix said:

Before you all start grabbing the pitchforks and firing up the burning torches, please bear in mind that these images could well be part of an A/B test.

I just packed away my pitchfork and put out my torch. Your detective skills are far more effective.

16 minutes ago, jagnje said:

if not there is one more hacked camera we can take clips of flowers and dogs with. 

My cat disagrees.

29 minutes ago, Andrew Reid said:

Has anybody come across any Fuji LUTs for F-LOG yet? I read there were some. However, such is the interest in the camera, 62 pages on this thread alone is a bit of a haystack to go looking in for LUTs

Fuji already gives you Eterna, I built my own LUTs for the rest. They are created for the X-T2, but they seem to work with the X-H1 as well. Remember to switch your clips to video range, that's what the LUTs expect.

8 minutes ago, DBounce said:

I think DNG or Prores would be better... and 12-bit. What kind of file sizes would 16-bit raw yield? 1 ~ 1.5 Gb/s give or take... that’s a lot of data. Even if it can.., and I believe it can, there are likely reasons why Samsung didn’t push it to these extremes. 

4K 16bit uncompressed RAW at 25 fps needs 422 MB/s. That's megabytes, not megabits. Based on the information we've been given so far the body can handle that, but you can't write that to the SD card, the fastest UHS-II cards can write up to 250-260MB/s continuosly. So you either compress RAW, or lower the number of bits.

46 minutes ago, Arikhan said:

For me (!) it's great.

It's great isn't it? :D

@mkabi He said earlier in this thread that it's internal, that why we discussed UHS-II capabilities.

Just now, Robert Collins said:

Well inherently pixel interpolation is a 'fudge' more than a 'fix' but camera companies are pretty good at it.....

Unfortunately it doesn't work all the time. I had only one hot pixel on my X-T2, but I could not remove it with the built in remapping function. Tried it multiple times.

1 minute ago, Robert Collins said:

Actually this is a pretty old fix for dead/faulty pixels on a Sony sensor. Basically any sensor will have a number of bad/dead pixels. I believe that QC limits the number to 1:10,000 (or something like that.) What the camera does when it sees a dead pixel is interpolate it.

What happens if you set the date well ahead of the actual date is that the camera searches for dead pixels and then remaps the sensor for any new dead pixels by interpolating any dead pixels it finds.

Oh I didn't know you have to do this procedure for a pixel remapping. On my X-T2 there was a menu called pixel mapping, so it was kinda obvious.

That was fast. If that's a true fix, that's one of the weirdest I've ever seen.

If you already own an A7III, you might want to check for this, the guy says he noticed it in different owners test shots as well:

Long but informative review:

This is just talk, noone is rushing anything. I don't even have an NX1 I for one trust Arikhan that he won't come here when he needs to be left alone. That's what I would do.

@Arikhan So the sensor is capable of capturing 28MP at 240fps, and the processor is capable of compressing and formatting it realtime. That's awesome, but we are talking about uncompressed RAW video right? Like Magic Lantern. I'm a bit confused now, because my question is still unanswered. If we have something like Magic Lantern RAW, but with 4K, how do you write that massive amount of data to an UHS-II card?

1 hour ago, Arikhan said:

@All - to say a little bit more...The speculations on data rate transfer limitations are not useful - because I've "opened the box"...Now I have partial direct access to the heart of this beast..

 

I believe the speculations ended with the conclusion that you are only limited by card speed. Did we get at least that part right?

21 minutes ago, Mokara said:

You can't. The fastest cards available will be able to just cope with 10 bit at most, and that is assuming you only use a 4K crop (in other words, no oversampling).

Okay, it seems like we all come to the same conclusion after all. Now we only have to wait and see.

1 minute ago, Matthew Hartman said:

My guess is it's doing 8/10 bits. But even with 8bit RAW, the image should be much more malleable in post and recover shadows and highlights better, you control the amount of denoising and don't have to worry about nailing perfect WB or using log. All this is very welcome over H.265 imo. 

But that was my conclusion earlier and Arikhan said I was completely wrong.