|
|
Post by turboxray on Jul 8, 2020 17:12:46 GMT
Thanks for breaking everything down. Since that custom palette file was for Mednafen which outputs RGB colors, I was curious how it manifested. I took a screenshot of Chris' 512 color demo and saw that each color is indeed unique and that colors are not currently in a consistent transition. So straight red and blue favor the dark end and never reach full RGB red or blue value. The brightest straight red also has a big chunk of blue in it. It will be interesting to see what the RGB equivalent of the final palette will look like, since we can only work in RGB color for pixelart. If you mean the girl's green hair in that DH screen cap, I can see a shade between the dark and light green on my phone in both. Was there a palette file posted here? It's probably pretty old/rough compared to where we're at now. There have dozen generated, tested, etc. We have a method for looking at it mathematically (the relationship mapped onto the color space), but we also test out against different games as well (perceptual test). If you're interested, I could PM you a more recent one for mednafen. I really don't want to post palettes here, because we're in the process of tweaking/checking/refining them. Phone's aren't exactly correct. I know that my iphone 11 is not. The color isn't completely invisible, but it's near invisible in straight RGB - the point where I would never use it for RGB for that close of range. That said, I realize that posting comparison pics, some people's monitors will have off/extreme/geeky gamma correction and such, and in some edge cases in RGB might make out the color difference slightly more. That said, if you have a capture card and would like to help out, send me a PM. Consumer capture cards aren't exactly accurate (especially if they compress to lower than YUV 4:4:4.. like most mpeg ones do), but I'd like to collect some different captures from cards to compare similarities. This is the current demo rom used for capturing (or for looking on your vector scope): gitlab.com/RickSugar/pcecolordemo |
|
|
|
Post by turboxray on Jul 8, 2020 17:16:55 GMT
I'm curious how this interacts with the TurboExpress; does that use the raw RGB or does it also go through this lookup table? If I were to guess, I'd say RGB. Simply because that screen technology wasn't even close to mature for a correct RGB ramp (though not as bad as the GameGear). Even the GBA has serious issues and that was like a decade later. There isn't really a need to preserve accuracy for a display that couldn't handle the accuracy. |
|
|
|
Post by Black_Tiger on Jul 8, 2020 18:16:50 GMT
I'm curious how this interacts with the TurboExpress; does that use the raw RGB or does it also go through this lookup table? Same with the LT. When I had one the sharp picture and vibrancy looked more like emulation color, but I wasn't looking for the odd coloring elements that stands out through stock composite. |
|
|
|
Post by Black_Tiger on Jul 8, 2020 18:23:02 GMT
Thanks for breaking everything down. Since that custom palette file was for Mednafen which outputs RGB colors, I was curious how it manifested. I took a screenshot of Chris' 512 color demo and saw that each color is indeed unique and that colors are not currently in a consistent transition. So straight red and blue favor the dark end and never reach full RGB red or blue value. The brightest straight red also has a big chunk of blue in it. It will be interesting to see what the RGB equivalent of the final palette will look like, since we can only work in RGB color for pixelart. If you mean the girl's green hair in that DH screen cap, I can see a shade between the dark and light green on my phone in both. Was there a palette file posted here? It's probably pretty old/rough compared to where we're at now. There have dozen generated, tested, etc. We have a method for looking at it mathematically (the relationship mapped onto the color space), but we also test out against different games as well (perceptual test). If you're interested, I could PM you a more recent one for mednafen. I really don't want to post palettes here, because we're in the process of tweaking/checking/refining them. Phone's aren't exactly correct. I know that my iphone 11 is not. The color isn't completely invisible, but it's near invisible in straight RGB - the point where I would never use it for RGB for that close of range. That said, I realize that posting comparison pics, some people's monitors will have off/extreme/geeky gamma correction and such, and in some edge cases in RGB might make out the color difference slightly more. That said, if you have a capture card and would like to help out, send me a PM. Consumer capture cards aren't exactly accurate (especially if they compress to lower than YUV 4:4:4.. like most mpeg ones do), but I'd like to collect some different captures from cards to compare similarities. This is the current demo rom used for capturing (or for looking on your vector scope): gitlab.com/RickSugar/pcecolordemoI was trying to capture RGB and composite/svideo examples earlier, but I don't trust my framemeister and capture devices. Sometimes when I capture video the color is completely skewed. I've been trying to find time to hook the framemeister directly to a monitor and test/calibrate it. The s-video mod on my TurboDuo looks the same as composite for color, but is a much clearer image. I'll try that demo out in the meantime anyway and see how it turns out. |
|
|
|
Post by dshadoff on Jul 8, 2020 18:40:56 GMT
It will be interesting to see what the RGB equivalent of the final palette will look like, since we can only work in RGB color for pixelart. The palette supplied here is already pretty close; perhaps 90% of the way there. Each of the candidate palettes at this point are either not as good as it, or reasonably challenging to say that they are an improvement. In any case, it seems like we're getting pretty close. If you mean the girl's green hair in that DH screen cap, I can see a shade between the dark and light green on my phone in both. Then you either have an interesting phone or really good eyes. I can't make out the difference for RGB on my desktop monitor at least (whereas the new palette is easy). |
|
|
|
Post by tatsuya79 on Jul 8, 2020 21:35:11 GMT
I found a case where the new palette makes the difference between some tints less obvious, in Legend of Hero Tonma in the blue background: |
|
|
|
Post by turboxray on Jul 8, 2020 22:32:01 GMT
I found a case where the new palette makes the difference between some tints less obvious, in Legend of Hero Tonma in the blue background: Which palette is that though? The one Dave posted 9 days ago or the latest WIP from the group? |
|
|
|
Post by tatsuya79 on Jul 8, 2020 23:04:08 GMT
The one posted by dshadoff on page 2, so 9 days ago.
I haven't found a more recent one.
|
|
|
|
Post by dshadoff on Jul 8, 2020 23:27:47 GMT
I haven't shared a more recent one.
After that one (which I stated at the time wasn't final), a few calculation blind alleys were explored which made the palette worse for a while - but I think it's now getting close to optimal. ...Close enough that a release should be imminent, so that the one we share should be the final one.
But don't forget as well, that you should be comparing against a TV or composite monitor, not RGB. The artist(s) may have chosen that palette for a reason.
And then, of course, there are some games which seem to have been developed for RGB (especially those ported from Amiga), which should probably be played with the RGB palette...
|
|
|
|
Post by Black_Tiger on Jul 9, 2020 1:00:23 GMT
It will be interesting to see what the RGB equivalent of the final palette will look like, since we can only work in RGB color for pixelart. The palette supplied here is already pretty close; perhaps 90% of the way there. Each of the candidate palettes at this point are either not as good as it, or reasonably challenging to say that they are an improvement. In any case, it seems like we're getting pretty close. If you mean the girl's green hair in that DH screen cap, I can see a shade between the dark and light green on my phone in both. Then you either have an interesting phone or really good eyes. I can't make out the difference for RGB on my desktop monitor at least (whereas the new palette is easy). It's a bit of both. I saw it even in the potato quality jpeg thumbail. But I've been porting over "impossible" higher bit pixelart to "PCE 36 integer" palette for many years now and work RGB values when doing original pixelart instead of picking from a color sheet. Both are more of a brownish green, but the one on the right blends a bit better. It's not seemingly invisible like that Startling Odyssey II sky and some other examples. |
|
|
|
Post by Black_Tiger on Jul 9, 2020 1:12:16 GMT
And then, of course, there are some games which seem to have been developed for RGB (especially those ported from Amiga), which should probably be played with the RGB palette... This is why I want to narrow down how inaccurate current RGB mods are and hopefully one day get something balanced like the current Mednafen 36 integer RGB values. I already have accurate color composite and s-video that looks the same. I'm also hoping that we can piece together an eventually complete list of games' ideal color mode display with notes for mixed cases. The Neo Geo ports are a good example where in emulation and RGB they look strikingly similar to the source material, which were developed for hardware supporting subtler shading, but the devs chose vibrant coloring. I'm guessing that games like LoXII, Golden Axe, Legendary Axe II, Neo Metal Fantasy, Metamor Jupitor, and Vasteel benefit a lot from stock composite. |
|
|
|
Post by dshadoff on Jul 9, 2020 2:25:43 GMT
The Neo Geo ports are a good example where in emulation and RGB they look strikingly similar to the source material, which were developed for hardware supporting subtler shading, but the devs chose vibrant coloring. I'm guessing that games like LoXII, Golden Axe, Legendary Axe II, Neo Metal Fantasy, Metamor Jupitor, and Vasteel benefit a lot from stock composite. I'm thinking that anything which was written first and foremost for the PC Engine (and not a port) will benefit from the composite palette. ...But there's another layer down the rabbit hole. We also saw on the Startling Odyssey II overworld (just before entering battle) that the grass has addiitonal colours when shown on a composite monitor than are shown when displaying using just the composite palette. The specific colour-transitions in the grass cause NTSC artifacting, which is used to great effect on this grass scene... points of darker colours which don't actually exist in the source material appear as part of the image, and complement it well. The artists definitely planned this aspect. We don't yet have a solution for that though... but we might start looking into it in the not-too-distant future. |
|
|
|
Post by turboxray on Jul 9, 2020 3:08:53 GMT
I'm also hoping that we can piece together an eventually complete list of games' ideal color mode display with notes for mixed cases. I mean that's definitely subjective, but also kind of irrelevant. This only about accuracy. And having the option for that accuracy - just not limited to an original composite signal (for obvious reasons). RGB was never an option. This isn't the Genesis RGB vs Composite debate. RGB only became an option after the fact, when people made RGB mods after the system was dead, as well as emulation. Your target set of colors, for what we experienced as gamers, was always from the VCE's YUV color space. Whether a dev or artists did a straight RGB down/upgrade for a port, or created original content - the delivery for those colors were always through the custom color set of the VCE. It should be close to invisible both mathematically and perceptually, but not completely. Again, a display that is inaccurate RGB ramp.. and it happens to fall on that threshold will make these low color issues stand out. I think it's a hard case to make to say that color, as a detail and coverage amount in her hair, is a valid choice. Shading wise, it doesn't really make a whole lot of sense (also, jpeg messes with colors because of lossless compression.. I though have posted PNGs. I was trying to get this board to image link to it, but this board's software is soo finicky). Do you mean as proof or for an artists tool? |
|
|
|
Post by spenoza on Jul 9, 2020 3:09:41 GMT
I can see the greens rather well on my XR. iPhones have pretty fantastic color reproduction on most models. I do see the difference in the hair
|
|
|
|
Post by turboxray on Jul 9, 2020 13:33:35 GMT
I can see the greens rather well on my XR. iPhones have pretty fantastic color reproduction on most models. I do see the difference in the hair Just to Note: the screen does, yes, but the software that controls the RGB transformations on you phone.. is not concerned with accuracy. They're concerned with "vibrancy" at the cost of accuracy - that has been the trend for a number of years now (Apple started it). It's no different than the "Red push" trend and general over saturation of HD sets starting in the 2000s. Not to mention that changing the backlight levels of the phone dynamically changes how it transforms colors for output (I don't mean analogue, I mean digital domain). I've seen the results pic results for all the WIP palettes we're creating, and see them on my iphone as well. And I'll check it on my phone (at different brightness levels). That sky gradient in question for SO2, is a very tricky color - I've seen my iphone push it in the violet side more than any other devices (for these WIP palette testings) but the rest of the blues are as drastic - and that changes depending on the brightness level on the phone for the exact same image. I have three 4k HDR sets.. all capable of displaying way more than 24bit color, yet one of those HDR sets has a RGB ramp that produces about 3 or 4 weird artifacts in the straight 512 RGB demo where the stepping isn't uniform for a few colors. That crazy because that should not even be a thing, but it just so happens those colors fall on a cusp of whatever correction threshold they're implementing on the set. I worked as a video engineer throughout the 2000s and had to deal with this crap constantly - and part of that was calibration of displays, and color profiles between different editing/pixel art/applications from different people working on the film production. It's the software developer's joke of "Well.. it works on my machine". |
|
