Personal View site logo
'Apocalypse Now' Experimental Series 1 Thread - BOOM, Intravenus - cbrandin/driftwood AN Soft/Cinema
  • 1089 Replies sorted by
  • Thankyou Chris for the explanation that I did request you write to avoid this kind of thing.

    For the record Lee, 'pseudo 444' was used to 'glamourise' the market launch of the AN setting but was title changed to the word 'simulate' very soon after. You took it very deeply. Well it certainly pulled you in. We're all walking around with 'hunches' that assumed it would soften the look of Pany lenses and that it proved correct.

    The 3x3 luma matrix works well. And we are noticing better chroma control for chroma red and chroma blue in the 4x4 part of the matrix.

    Changing the values of the luma 3x3 original values into a 4x4 matrix has proved interesting and one that is still 'out for results'.

  • All I did was come up with a new matrix idea. It's fine with me if people expand on the idea. In fact, I think it's great because I don't have the time to build settings these days. My thanks goes out to all of you who do!

  • I did some sharpness measurements for the soft 444 matrix versus the factory matrix using sharp Panasonic glass at an optimal aperture. The soft 444 matrix has a resolution of about 860, and the factory matrix has a resolution of about 920. That's about a 7% loss in resolution - which isn't very much. The reason for this is that the CODEC in the GH2 doesn't actually resolve the theoretical limit for 1080p. What you get in return for that 7% loss is much better gradations and color rendering.

    It's hard to characterize these color subsampling models as 4:2:0, or 4:4:4, or 3:3:0 because those color model numbers don't really relate to how color is sampled in any sensible way. 4:4:4 is where luma and colors are rendered the same, 4:2:2 is where there are twice as many luma pixels as chroma pixels per channel, and 4:2:0 is where there are four times as many luma pixels as chroma pixels per channel. So the numbers don't really work in an intuitive way.

    What I discovered is that because of the way the GH2 codec works there is a loss of resolution (vs the theoretical limit) in the luma channel, but not in the chroma channels. By using a 3x3 matrix for luma, and 4x4 matrices for chroma they actually come out closer in resolution than theoretical calculations would indicate.

  • @driftwood - Please calm down, you don't have to take this discussion as a personal attack. You've released six different versions of Apocalypse Now, four with Chris' pseudo-444 matrix, and two more with a new "sharp2" matrix. Some of us found that confusing and I took some time to figure out which was which and explain the differences in precise detail. And to clear up any confusion on your part, here are a few answers to your questions:

    "The 444 Soft matrix IS INDEED Chris's Matrix setting"

    Yep, that's what I explained above.

    "[sharp2] should have been the original SHARP release - a edited version of chris's base settings.

    Fine, but your sharp2 matrix no longer embodies Chris' 3x3 luma concept that made it a pseudo-444 matrix. I had to investigate this in detail in order to determine which Apocalypse Now version to use to evaluate Chris' original concept.

    "Do you think we're miscreditting him?

    No, you've made it clear that the Soft Variant is Chris' pseudo-444 matrix and the Sharp2 Variant is yours.

    Honestly Lee I dont know whats got into you lately?

    It's perplexing trying to figure out which variant of the latest matrix does what in which mode.

    Do you know what pain is?

    As a matter of fact, I do, but my physical therapist says that's all part of the healing process.

    How bout you get off your little high horse and get cracking on a flowmo version?

    Last time I checked, Flow Motion v2 was still working fine...

  • I finally get in after a long day travel to France and switch on the lappie to see @lpowell here once again determined to have his moment of glory.

    So let me clear up a few things first.

    Chris told all of us his mattix ideas - he probably told u at the same time - so fuckin what? We went out and put them to the test. Do u want me to publish mine and Chris's conversation?

    The 444 Soft matrix IS INDEED Chris's Matrix setting - the p and b frames are from our work.WHEN did we EVER state otherwise?

    Chris's settings have then be elaborated, experimented (call it what you what u will) to produce tad sharper settings based on research & testing from use with some of my soft lenses and then recommended as tryouts as beta candidates. This should have been the original SHARP release - a edited version of chris's base settings. Thats why I still credited the man who I respect very much.

    He came to you LAST July or this July just gone? Did he talk to you first? Who fuckin cares? The AN settings tested by our beta team which included Chris's prescence & certainly participation. Does that make you feel better? Do you think we're miscreditting him?

    Honestly Lee I dont know whats got into you lately?

    Is it hurting you that we put out some settings with this matrix? Besides, its a good mattix for a certain job, but its NOT neccesarily the be all and end all. Its just another setting (not patch btw)

    Finally we (the beta team) have spent a lot of time developing and testing settings - Im talking fucking days of testing here (do you know what pain is?) so how bout you get off your little high horse and get cracking on a flowmo version? Or I can make you one if you like?

    Rant over There I feel better now Lee :-). Back to my holiday. No hard feelings mate, but you really are getting to be annoying.

    Ill reply to questions asked here in my own time. I dont NEED you to step in before I even get a chance. If you wanna wade into quant stuff lets do it. I certainly would love some answers about one or two of your numbers in Flowmo.

    Wanna play?

    bkmwcd's 3 GOP 24p renders better than Flowmo - you should give it a try :-). Thats with my testing and its MY personal view. Thats all.

    Also Ive tested my settings against Flowmo and Im very very happy with Nebula and DRewnet. Very happy.

  • @ValentinDeluy Yeah that would be great send it to my email dedixon@live.com

  • @peternap Yeap. I think this new setting is quite interesting. Less perfection. More character. Wabi-sabi :)

  • @lpowell Thanks for the explanation. Finer chroma channels might give more perceived color information though... at the expense of losing luma resolution.

    @plasmasmp That's a good plan.

  • @stonebat People like me who mix Lumix lenses and MF lenses are in conundrum.

    I couldn't agree more!

    As usual though, I'm going back and forth. Flow Motion 2.2 is my does all with all....setting. It's not necessarily cinematic, but neither am I.

    I have to admit though, that what I've done so far with Apocalypse Drewnet soft is exceptional, at least to me. There's something special about it.

    Lots of comparing to do.

  • I certainly get what the encoder is doing. But are we losing real detail, or getting closer to what the video mode should actually look like?

    I haven't seen an actual diagram how the GH2 processes 15.9MP down to 2.1MP 1080p. The method obviously isn't perfect since there is still luma aliasing, but it is noticeably better than the alternatives. Even with manual lenses there is still a hint of sharpening that can be seen more severely when comparing to other cameras. The zacuto shootout really points those issues on high contrast surfaces when you compare it to higher end cameras.

    99% of the time I use manual lenses. Right now I'm just testing the ol Lumix lenses that have been sitting in storage. If the image is unacceptable with manual lenses I'll switch back.

  • If I have extra Gh2, I'd have it with this new setting for Lumix lens only...

  • @plasmasmp @stonebat What the AVCHD encoder is actually doing is recording the luma channel at 1920x1080 and the two chroma channels at 960x540. To fully emulate a 4:4:4 color space, you'd need to downsample the luma channel to 960x540 and encode all three channels at the same quality. Athough this can't be actually be done on the GH2, Chris' 3x3 matrix does eliminate most of the finest detail from the luma channel, which he figured would degrade the luma resolution to somewhere between 1080 and 720 lines.

    Chris' matrix doesn't really add "50% more color information", it just encodes the chroma channels as finely as possible, while eliminating the finest details from the luma channel to take its resolution down closer to the chroma channels. With over-sharpened Lumix lenses, this might well produce a look that you prefer for its softness.

  • The setting would add 50% more color information, and that might be useful in certain situation. People like me who mix Lumix lenses and MF lenses are in conundrum.

  • I don't think it looks anywhere near as low as 960x540. I've been testing with cbrandin's settings because I want to use the la7200 with Panasonic lenses without sharpening. Here's a still from the 4:4:4 GOP6 Soft.

    20mm + la7200 (and the LA7200 is an extremely soft anamorphic)

    1080p.jpg
    1920 x 797 - 249K
  • @peternap No, I think the results from Apocalypse Now have confirmed my hunch that Chris' pseudo-444 matrix would produce soft images.

  • @Mirrorkisser

    Pseudo sounds rather negative.

    Just part of the pseudo-branding hyperbole we're getting with settings. I think that in this hi-commercial world we're getting pushed to find names which get noticed and remembered.

    To this end, Pseudo 444 fills the bill, but has the unwanted connotations we both mention.

    AT the expense of creative freedom, I'd like to see something like a nomenclature for each settings family, plus an alphabetized nickname for each major increment, like Canonical does with:

      Ubuntu
    • Dapper Drake
    • Edgy Eft)
    • Feisty Fawn
    • Gutsy Gibbon, etc
    Add to this a standard Version Control and we've gone full-conservative. (Anywhere in-between would suit me). image
  • @lpowell I was curious about any side effect from losing half of luma info, too.

  • .@LPowell Are you going to try to incorporate Chris's matrix?

  • @ValentinDeluy Awesome stuff, amazing footage !!

  • @ValentinDeluy chouette mes compren ke dalle! Parole stp?

  • @ValentinDeluy That's some sick editing! Great music video btw! Those guys are great!

  • Thx @shian i was thinking the same...
    Its in the deblocking table, i change it, in my setting, and now it looks better..
    But i have still a lot off testing to do....

  • @Tobsen Two tests with Valkyrie 444 soft HBR mode 25fps (Pal). First stopped on its own initiative at +/- 23 minutes (2 times 4.29GB + a bit more in a third file) and second stopped at the very end of the second 4.29GB at +/- 18 minutes. Not really spanning, at the end !

  • @LPowell that would explain the appearance of noise in the green channel, mostly prevalent on green objects which I had never seen before. I thought I had done something wrong, but apparently it's the patch.

  • @lpowell: thanks for your reply. I am just looking for the hack that looks the least videoish. The hype and quest for super duper shallow depth of field is a bit overdone in my opinion, but videoish is even worse. I like flow motion a lot, but i am always curious to try out new things and also like some of nick's settings.

    Which patches do you like the most for their cinematic (silly word, because cinematic can be a lot of things, but i guess you know what i mean, i mean it in the sense its used in this forum the most) looks?

    Cheers

This topic is closed.
← All Discussions