Jump to content

Sage

Members
  • Posts

    659
  • Joined

  • Last visited

Posts posted by Sage

  1. 1 hour ago, deezid said:

    2.5K looks bad on 4K screens though, lots of smear and macro blocking especially on YouTube - and you can still select streaming quality manually. ;)

    Or chrome which can be color managed in chrome via about:flags url: Force color profile

    Select sRGB (standard screens) or DCI-P3 (Mac and OLED) to enable color transform.


    Still a bit pink as Alex already mentioned.

    Righto, I like to force 2.5K for motion and load speed. But I know you're a 4K guy :)

  2. 1 hour ago, superporpoise said:

     

    @Sage do you have a recommendation for bitrate for 2560x1440 h.265 for YouTube?

    It appears there is only the option to set Quality to Auto-Best, or to restrict the upper bitrate (unlike Premiere). I gather that they've looked at the needed bitrate for each quality ranking, and kept the works 'under the hood'

  3. When exporting to YT and Vimeo from Resolve, I've found h.265 to be optimal. There is a slight magenta shift when the video sites re-encode the footage, and I've found it to be slightly less with h.265 (the colors are slightly more accurate). The difference for h.265, to compensate during export, is a node with -3 Tint. H.265 is also a much smaller filesize, which is less heavy on the Vimeo upload quota, and there is no size penalty for higher resolution.

    However, I've found the sweet spot for streaming video sites is 2560x1440 (or 2560 pixels wide). The load times are not much worse than 1080p, but the apparent visual quality is much better (because 1080p is very compressed for streaming on these sites). The load times for 4k are much higher, with less of a leap in visual quality. Also, the fluidity of playback for 2560 is smoother than compressed 4k for streaming.

    Here are my Resolve settings (horizontal resolution to match aspect ratio):

    export.thumb.PNG.5309cd1262b7b73d02c7be167f3b4152.PNG

  4. 8 hours ago, RCV said:

    Can someone link the p4k thread.  Perhaps,  there's more footage. 

    Great tutorial btw, i have it bookmarked

    Righto, this is the one:

    I intend to revise the tutorial to be more concise, yet more informative and clear.

  5. 1 hour ago, crevice said:

    Alex - would I be able to use the pocket 4K version on the URSA mini pro, if I used a color space transform to get to the pocket 4K first. Or does that not make any sense? I’m very eager to see how good URSA mini footage would look with this. 

    That could be done; granted it wouldn't be tailored to the sensor. I'd like to support it directly, and I think there may be some rental options that would make that feasible.

  6. 24 minutes ago, katlis said:

    @Sage Just checked the GHa thread and noticed you released this P4k version, awesome! I just grabbed a P6k last month and have been excited about this. I know you're still working on that version, but in the meantime I'll test it out on the 6k for fun. From what you've seen using the P4ka on P6k footage, any tips on how to make the most of it or get it as close as it should be in your eyes? 

    I haven't looked at it in-depth yet, but from what I've seen there are subtle differences (i.e. green). Overall, the default color of the 6k is a little nicer. The sensor response as things fall into shadow is a little different. I'll know more soon.

  7. 4 hours ago, Zeng said:

    Amazing job. Wish you could do it for more cameras. I could provide you with samples of most of them side by side with Alexa, and you just do the math. How about it? :).

    Thank you Zeng; I need the cameras in hand, as the measurement phase is the trickiest, now that Gen4 is figured out. The code is based around a sample system I made that has ~1k samples per slice, with ~40 slices, and the measurement conditions are defined; often it must be redone several times before everything is sorted out with a new camera. I will move with considerably more efficiency to support new cameras, now that Gen 4 is finalized.

  8. 3 hours ago, majoraxis said:

    @Sage My question is if one plans to do further tweaking like color correction, film emulation, reduction for contract and saturation, would it be better to have variations of the EC LUT to accomplish these tweaks because doing it in a single conversion (using just one EC LUT instance) would turn out better than using the EC LUT and following it with more nodes to tweak the look. Thanks!

    Oh, I understand now.  EC Gen.4 is at the theoretical limit of smoothness, acting like a pass through of source material. This means that if the source material is of sufficient bit depth (for sparse regions), its like the conversion is transparent. This allows for Pre format stacking to support formats like HLG and Cine-D (which will see a major improvement with Gen4) - two V4 cubes stacked is identical to one summed, unless they are intricate 33x with bad interpolation (AE - bad, Premiere - good, Resolve with tetrahedral - very good). Nodes following thereafter may introduce artifacts of their own (the final space is only as clean as the most problematic link in the chain), but EC as yet doesn't cover creative looks etc. One may use the EC gradient chart to verify that each new grade step is not introducing problems in the wider space.

    @Mark Romero 2 Thank you - what camera do you use?

  9. 59 minutes ago, majoraxis said:

    Also,  is there the concept of having different EC LUT versions that are tuned for high and low dynamic range scenes as well as the standard EC LUT?  Or maybe this does not apply and only one EU daylight LUT is need for any amount of dynamic range in an daylight light scene?

    That's right, only one is needed for the entirety of the range. Conversion accuracy is 'exposure independent'; though white balance will affect accuracy. With Exp Comps, it is now possible to dynamically allocate scene range for the 709 display variations in post without affecting core accuracy.

    Spell this one out for me - 'So would having a set of EC LUTS for emulation film stock or an ideal sensor response LUT would benefit from having an EC LUT for each target output to do it in a single conversion?'  I couldn't quite follow

  10. @jack jin Sure; you'll find the P4K to be similar. The summary of it is that Day will be greener under tungsten (always ensure skintone maintains contact with the skintone line on the vectorscope afterwards). The whole color space is slightly different, with each color weaving a divergent path up through the space.

    @majoraxis Thanks majoraxis, all things are equal up to the cube; ABL must follow the conversion node, and is only on P4Ka for the comparison. WB is actually on the raw clip itself rather than a node (unlike GHa), as the Camera Raw tab is used for a completely accurate WB adjustment

    @MeanRevert These are good notes for my next tutorial revision :) The Pre and Posts are much better - Exp Comp Pres are actually complete color conversions that capture the difference in sensor response with varying brightness (they include the rolloff and brighter than 1.0 region as well). And they are incredibly precise to actually stopping up or down in camera (preserving core LogC accuracy). The Posts simply have hue lock built in, and are an optimal luminance adjustment (preserving linearity above black). They are really fast to use, and are a major time saver. They also have a negative region preserved (below 0.0), and therefore not destructive to dark saturation data. That said, for a film, additional luminance adjustment after Post is recommended, to conform the scene details to cinema standard scene luminance (supplemental waveforms)

  11. 55 minutes ago, RCV said:

    this is from the genius Sage.

    Brilliant work. cant wait to use this on the p4k shoots I have lined up.

    Sage has to do a pipeline tutorial because for some reason, folks arent following the pdf

    Ah, nice find. I was going to post that in the P4K thread - though I will be posting the tutorial video in the P4K thread ;)

    Its true; almost everyone doesn't read the Pdf in the first week, and then maybe a few later on. So I've come to the conclusion that a video and quick guide are very much needed

  12. 25 minutes ago, RCV said:

    This is lovely news. wondering why u didnt post about this before. Welcome the new changes implemented. This camera will absolutely shine. Just bought the new portkeys monitor to go with my p4k which has been doing wonders on few gigs. I am excited to use it on my feature shooting 2020. The GHalex has served me well

    Wish there were more examples of this with the p4k but I know there will be a lot coming out soon. 

    Sage, you are the main man

    Thanks RCV! I've been absurdly busy. One of the standout features of the P4K for me (coming from the 5) is being able to preview 33x on the big screen. Its so much better than the vlt. But IBIS is sorely missed. I wanted to shoot an epic creative piece, but alas was too booked, and the sunny season slipped by

  13. 47 minutes ago, Video Hummus said:

    Hopefully we will see a Gen.4 for GH5S. Great work these look incredibly similar. 

    Thanks Video Hummus; the S will be Gen.4 right out of the gate, and that is the very next camera (along with the GH5 update)

    21 minutes ago, zerocool22 said:

    So if you work with these, you would get the same(similar) starting point as alexa footage(without the DR)? But if you throw in some Alexa luts afterwards will they clip on the pocket 4K?  

    The core conversion is to a measured LogC (what Alexa records to). Then there are the display variations, which include the 709 display primaries from Arri all-in-one. Then there are the display look-ups from Arri itself (to be added to LogC), and their Rec 709 does clip far gamut.

×
×
  • Create New...