caseywilsondp

Rod Gotfried : https://vimeo.com/157995688
https://vimeo.com/191130528

Joshua Morin:

https://vimeo.com/159937841
https://vimeo.com/118014639
https://vimeo.com/113397445

Casey Wilson:

https://www.youtube.com/watch?v=pIzhGCzBtUQ
https://www.youtube.com/watch?v=lY0D-oocp7s

Oliver Astrogolo

https://vimeo.com/181612110
https://vimeo.com/126747807
https://vimeo.com/137925379

Bringing this back to life, this time with a picture :

Honestly - I never thought the NX1 hack would get this far!

I can remember long ago, when we were all but children playing in the garden, back when Kidzrevil first posted this challenge....

At the rate some of the folks here are carving deeper into the guts, I wouldn't be all that surprised if one day someone DID find a way to get a RAW video feed out of these kittens of ours.

In the meantime it sure seems to be a resounding success even just in it's first couple of months, with the increased bit-rates cleaning up some of the problem areas in just the way I expected it would. Kudos to all involved.

Now if only Otto or someone else could write a sub-script that would prevent me from pointing my NX1 at bad subject material...

I feel colors end up looking very natural:
My testing results:
http://caseywilsondp.com/2016/01/27/best-nx1-settings-for-4k-video/
Luts for nx1:
http://caseywilsondp.com/2016/03/24/samsung-nx1-free-lut-pack/
 

I feel colors end up looking very natural:
My testing results:
http://caseywilsondp.com/2016/01/27/best-nx1-settings-for-4k-video/
Luts for nx1:
http://caseywilsondp.com/2016/03/24/samsung-nx1-free-lut-pack/
 

Thanks for making theses available!

I feel colors end up looking very natural:
My testing results:
http://caseywilsondp.com/2016/01/27/best-nx1-settings-for-4k-video/
Luts for nx1:
http://caseywilsondp.com/2016/03/24/samsung-nx1-free-lut-pack/
 

Time to make this topic alive again, comon people, lets get sharing !
 
 

ok just a quick preview of the test i shot earlier.
here are unedited 14 bit jpegs from the timeline. one is at bit rate 80mbps, 180mpbs, and one is 660+ mbps (shogun). can you tell from the stills (shogun might be obvious)? i'll upload the actual video files a little later. if anyone wanted to take a stab at grading these, i'd love to see it.



wanted to test banding, shadow detail/macroblocking, skin tones, but had to run the test really quickly so I chose these settings:
normal gamma, -10 contrast (to induce banding), -10 sharpness, 16-235
the color chart is just for reference, btw.
 

Cross post from dpreview:
As promised I was working on alternative way to run scripts off the SD card without entering the factory mode (as it limits several functions important to me).
Full text here
https://github.com/ottokiksmaler/nx500_nx1_modding/blob/master/Running_scripts_without_factory_mode.md
What it does: it modifies your root filesystem in camera so it starts a shell script from SD card (if present) every time you start WiFi.
Yes, it's rough. This is a first release just to enable everybody else to test and play around.
Touchscreen works.
Bluetooth works.
Camera works faster than in factory mode.
This might help with increased bitrate as well (don't know, can't test).
Have fun

Here is comparison of hacked and non hacked 120 fps (200% zoom). There is quite big difference in terms of compression and details.
Problem is that hacked is just 106 fps according Premiere...
First is hacked and second normal:
 

vasile, thanks for your hard work! this is exciting stuff.
did some initial tests of 80 vs 180 (highest my card can write continuously) and I see little to no difference, so double checked the file and sure enough it is 180mbps. That was just a rough shadow/highlight/detail test. Later I'll sit down with some lights and a person and set up a scene to really test it out. My guess right now is that hevc is extremely efficient, so the difference between 80 and 180 are going to be negligible. 320 might be different, but I can't test that until I get a faster card. This does seem to make sense because even at 660+ mbps to an external recorder, the main improvement is only in shadow macro blocking... even with nearly 600 more mbps. It's also being written as prores, so that could even account for some of the improvement.

I have mentioned this previously on other forums, the H.265 codec profile used within the NX1 is limited to 160,000 kbps, it's MAIN - Level 5.1, to go higher you need to change the codec profile. To get to 10bit the NX1/NX500 needs the newer version of the codec without which it is impossible.
https://en.wikipedia.org/wiki/High_Efficiency_Video_Coding

here are two scenes at 60p
normal gamma, -5 contrast, -10 sharpness. 16-50S lens clean (no filters)
first detail (detail on bush, as well as how the out-of-focus bit looks i.e banding)
int:

ext:

int 200% crop:

ext 200% crop:

as a still image, the full res images appear pretty similar, with the ext appearing to be slightly sharper. go to the cropped images, and we can see why. blocks of pixels around the edges of the bush are a bit larger/more noticable, and make the bush appear just a bit softer, and give the out-of-focus area a bit of banding.
here is a shot that can give us a good idea of the macro blocking that tens to occur in shadows:
int:

ext:

int 200% crop:

ext 200% crop:

externally and internally results are similar to before, you can notice more macro blocking in the shadows on the internally recorded image, and there is a very, very slight edge to the external image in highlight recovery (again the jpeg would suggest different, and i'm not sure why the export crushes the highlights like that, but check the original files).
*HOWEVER*
look at the external image crop... in the shadows... there appears to be a very fine fixed grid pattern (when they're bumped up significantly). it is very odd. now, you probably shouldn't expect to use any detail that's this heavy in the shadows (+150 via lumetri), but even so I'm not sure this grid pattern is much preferable to the heavy macroblocking of the internal image.
as a still image, i'd actually prefer the internal image, however with motion the macro blocking becomes very noticeable and the external seems more like noise.
ultimately, in motion, i find the external image to be a bit better, but if it's better enough to warrant an extra 1200-2000 (depending on recorder) is up to you.
full files:
https://drive.google.com/folderview?id=0BwO55ILG1OL-YnNqbGJDejhIcjQ&usp=sharing
 
 

We will be expecting eagerly your hacking/moding advances, until then just keep it humble, so to keep this thread useful and without any grey noise.

Please reread the whole thread, the nx1 and nx500 cpus are not the same at all. They reuse a lot of code, Parts of the encoding blocks reference nx300 things, it doesnt mean they are the same as the nx300. The architecture may be similar but from what I have seen code wise is quite a lot of difference, yes both are tizen, both have very similar control file structures, but look into them and the function calls change for obvious reasons, now Im not saying you cant use code from either, just that like everything it would have to be adapted. and modified with the proper nx1/nx500 references.

Some new stuff im working on :
 



All videoframes, im thinking of probably starting that stock footage thing sometime near.

not 120, but here is 60fps 1080p internal vs high bit rate
http://www.eoshd.com/comments/topic/18661-your-ideal-nx1-settings/?do=findComment&comment=129594

 
I'm not entirely sure the camera is outputting 422. 8bit 422 on other dslrs produce much less banding which leads me to believe the nx1 is outputting 8bit 420 over hdmi. 
macroblocking is precisely what increased bitrate will help us with, however.
80mbps:

600 mbps:

clearly here is some jpeg compression from the upload and its a major crop of the origional, but the difference i think is clear.
higher bit rates = less macroblocking and more fine detail.
it will not significantly reduce banding, we need 422 and ideally 10bit+ for that.

not 120, but here is 60fps 1080p internal vs high bit rate
http://www.eoshd.com/comments/topic/18661-your-ideal-nx1-settings/?do=findComment&comment=129594

 
I'm not entirely sure the camera is outputting 422. 8bit 422 on other dslrs produce much less banding which leads me to believe the nx1 is outputting 8bit 420 over hdmi. 
macroblocking is precisely what increased bitrate will help us with, however.
80mbps:

600 mbps:

clearly here is some jpeg compression from the upload and its a major crop of the origional, but the difference i think is clear.
higher bit rates = less macroblocking and more fine detail.
it will not significantly reduce banding, we need 422 and ideally 10bit+ for that.

Unfortunately it'll look like this:

colored and punched into problem area:

600+ mbps prores hq recorded on shogun from the nx1
of course that is gamma dr -10 contrast... so about as extreme as you can get, and maybe its usable in cases where you wont have gradients, but as i demonstrated a lot in the "your ideal nx1 settings" thread, increased bitrate will help us with shadows and highlights (we can improve usable DR by 2 stops is my guess), but we need 422 to get us away from banding.

Unfortunately it'll look like this:

colored and punched into problem area:

600+ mbps prores hq recorded on shogun from the nx1
of course that is gamma dr -10 contrast... so about as extreme as you can get, and maybe its usable in cases where you wont have gradients, but as i demonstrated a lot in the "your ideal nx1 settings" thread, increased bitrate will help us with shadows and highlights (we can improve usable DR by 2 stops is my guess), but we need 422 to get us away from banding.

Unfortunately it'll look like this:

colored and punched into problem area:

600+ mbps prores hq recorded on shogun from the nx1
of course that is gamma dr -10 contrast... so about as extreme as you can get, and maybe its usable in cases where you wont have gradients, but as i demonstrated a lot in the "your ideal nx1 settings" thread, increased bitrate will help us with shadows and highlights (we can improve usable DR by 2 stops is my guess), but we need 422 to get us away from banding.

Good news, bad news.

Good news is that I have learned a lot about the srp( samsung reconfigureable processor) architecture. Bad news. It is very very very very complicated. Machine compiled c into binary. Unknown yet when it is loaded into the firmware and if the .cpp files that are listed are infact that, or just other hard ware controls

So the srp is like an fpga in the way you can change what it does on the fly. It differs in the way that it uses VLIW instead of bits to do its magic.
Reverse engineering the binary into something workable.. is possible, but time will tell if and how much can be changed. The plus side of a simple isp changing values in softwear can do something like what the gh2 hack did. The downside being you are stuck with the hardware.
The plus side of the srp implementation.. Building a custom algorithm to..exchange the hevc codec with something else.. Could be possible. Unknown yet if the nr is a part of this or just a software implementation.

Also if the rumors were true about the ml guys approaching samsung, this would be why it didnt happen. They dont release the srp compiler. One instance of them giving access to it at the SAIT school in Korea is all I could find. very very proprietary.

But the dream for me of having a cheap 14bit 6.5k red competitor is what keeps me going. The confirmation of the UHS-II bus sweetens the deal. Going to investigate the possibility of a SD-SATA converter.. I wouldnt mind having a 1tb ssd attached to the back of the camera.


TL:DR This camera is complex and things are not getting easier but still progression is being made.

Ha, yes, we had some good interaction in that thread. And yes, for h265 you are probably correct. Prores on the shogun seems to care little between 0-255 and 16-235 once you actually grade the files. Sorry for this digression in this thread!
My main point earlier that I may or may not have made, is that increased bitrate will help us, particularly in h265 with things like macroblocking, unfortunately I doubt it will help us with banding (as 220mpbs+ already doesn't seem to get rid of it ala the shogun). while h265 is a very impressive codec, I'd really like to see something come along in the hack that will give us 422 (i really dont think hdmi out is 422 on this camera, but i have no evidence other than the terrible banding that still occurs). with gammadr and -10 contrast the camera is capable of a ton of latitude, it just can't be pieced back together in color without significant banding.

Ha, yes, we had some good interaction in that thread. And yes, for h265 you are probably correct. Prores on the shogun seems to care little between 0-255 and 16-235 once you actually grade the files. Sorry for this digression in this thread!
My main point earlier that I may or may not have made, is that increased bitrate will help us, particularly in h265 with things like macroblocking, unfortunately I doubt it will help us with banding (as 220mpbs+ already doesn't seem to get rid of it ala the shogun). while h265 is a very impressive codec, I'd really like to see something come along in the hack that will give us 422 (i really dont think hdmi out is 422 on this camera, but i have no evidence other than the terrible banding that still occurs). with gammadr and -10 contrast the camera is capable of a ton of latitude, it just can't be pieced back together in color without significant banding.