Quantcast

Page 3 of 14 FirstFirst 123456713 ... LastLast
Results 31 to 45 of 204

Thread: Best And worst Quality FMV (Streaming video) on the Sega CD?

  1. #31
    Mastering your Systems Shining Hero TmEE's Avatar
    Join Date
    Oct 2007
    Location
    Estonia, Rapla City
    Age
    30
    Posts
    10,095
    Rep Power
    111

    Default

    Regarding Sonic CD, it was never meant to have any compression.... and the timing is not based on VBLs or anything BUT the 150KB/sec CD data rate.... The video is in 8 frame chunks, each chunk is made of 32KByte sound block (1 second of 32KHz sign-magnitude samples), then there's 16x 32 byte palettes for each frame and the rest is 8 frames of uncompressed and unoptimized tile data...
    Death To MP3, :3
    Mida sa loed ? Nagunii aru ei saa "Gnirts test is a shit" New and growing website of total jawusumness !
    If any of my images in my posts no longer work you can find them in "FileDen Dump" on my site ^

  2. #32
    ESWAT Veteran Da_Shocker's Avatar
    Join Date
    Apr 2009
    Location
    Cashville,TN
    Age
    39
    Posts
    5,254
    Rep Power
    71

    Default

    Prize Fighter?


  3. #33
    Sega Nerd WCPO Agent oldskool's Avatar
    Join Date
    Jun 2010
    Location
    Las Vegas
    Age
    43
    Posts
    825
    Rep Power
    17

    Default

    Has there ever been any FMV home brew attempts, maybe something with hardware trickery/expanded palette/resolutions? I'd like to see the BEST it can do even if it's not truly a game. Anyone ever attempt to convert video into a Sega CD happy format? Cinepak or whatever?
    MY SEGAS
    [Model 1 Sega CD x 5] [Model 1 HD Genesis w/Component Video] [Model 2 Genesis] [32X CIB] [Master System Model 3010] [Nomad w/LCD mod] [Dreamcast] [Saturn] [Game Gear]

  4. #34
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    Quote Originally Posted by oldskool View Post
    A lot of this seems very technical to me, but one thing that I have noticed is that - if they would have digitally remastered every single frame it would have looked much better.
    I wouldn't doubt that they did manually edit some bad frames. But there's only so much you can do. And of thousands of frames, I'd hate to be the man to do it (from what you're suggesting).


    I think most of the FMV was just a rush job and they went with it. I've noticed colors where they shouldn't be and lot's of grainy graphics.
    It's a limitation of the video setup. Genesis is a tile display system. It doesn't lend well for video or even stills. But at least with stills, you have more time to work the image to make it look better. The wrong colors in certain areas are conversion artifacts. Errors. Colors that couldn't be represented had to be associated to something else (preferably close). There are actually a ton of conversion errors, you just notice the bad ones. The graininess is do to the dithering process. It also helps hide errors as well as help present colors that are possibly in a 8x8 graphic block of the screen, or even colors that don't even exist in the Genesis 512 color main palette.

    If you watch the Sonic CD intro, the colors are very clear and there is not a lot of grain going on, but something like Wirehead or anything else and it looks grainy. If they would have went frame by frame and touched things up, removed the grains from the graphics (so that there were one solid color) I think it would have looked much better.
    For animation, sure. But not for real life video. Although for some animation, dithering can be fine.

    Has there ever been any FMV home brew attempts, maybe something with hardware trickery/expanded palette/resolutions? I'd like to see the BEST it can do even if it's not truly a game. Anyone ever attempt to convert video into a Sega CD happy format? Cinepak or whatever?
    If it was simple enough to do, then it would have been done by now. Making uncompressed video and/or layered plane video is doable - for low color stuff. But add compression into the mix and optimizing for high color output, and you've got yourself one headache of a situation. It's complex to say the least.

  5. #35
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    31
    Posts
    9,725
    Rep Power
    64

    Default

    Quote Originally Posted by cj iwakura View Post
    Wirehead had the full screen FMV, though it was probably grainy beyond compare.
    Grainy or blocky? Graininess (dithering) should be unrelated to screen size, but compression artifacts would be.

    If they managed a decent framerate and no screen tearing at full screen that would be rather impressive both in context of compression and the VDP/VRAM limitations (though I'd assume it was 256x224, which would still allow double buffering, so DMA bandwidth would be the possible area of contention)
    It's probably not running faster than 12 fps to stay within the DMA limits of H32 (256 wide res mode) with screen only clipped to 224 lines. (games with boarders filling the screen lose the advantage of added DMA bandwidth from clipping -Night Trap and Sewer Shark even fill the full 240 lines for some odd reason)
    Assuming 256x224 at 12 fps, that's probably ~2.6-3:1 compression (depending on how much goes to audio and if mode 1 or 2 data is used -150 kB/s vs 174 kB/s) That's not THAT much more than Road Rash (256x152 15 FPS -higher in certain sections), only about 17-18% more compression.

    It does look pretty good on Youtube though:
    http://www.youtube.com/watch?v=U9hHdFbGevg
    U9hHdFbGevg
    And honestly, a good option is probably to crop the boarder all around due to overscan: something like 240x208 would fill so much of the screen (NTSC) that you'd likely not notice it too much. (plus you could manage 20 FPS before hitting the VDP limit with that much clipping -the issue would be if you could manage that much compression without too much degridation)


    Quote Originally Posted by TmEE View Post
    Regarding Sonic CD, it was never meant to have any compression.... and the timing is not based on VBLs or anything BUT the 150KB/sec CD data rate.... The video is in 8 frame chunks, each chunk is made of 32KByte sound block (1 second of 32KHz sign-magnitude samples), then there's 16x 32 byte palettes for each frame and the rest is 8 frames of uncompressed and unoptimized tile data...
    Yes, I never argued such. As Chilly willy mentioned in the 32x video CD thread, uncompressed video was a common format used ad is the case for Sewer Shark and Night Trap as well. (the framerate is double that of Sonic CD due to the lower resolution and lower quality Audio I'd assume)

    What I was arguing is that had it used compreasion, it could have looked a good bit better and included the full animation rather than doing the odd thing where they cut out wole sections of the animation and slowed the progression down (I assume due to it looking poor at 8FPS with more frames of animation cut in general -not so with a higher framerate)
    Likewise, using a smaller screen size would have facilitated such as well, like Sewer Shark or Night Trap (the dithering would be entirely optional, of course).

    There are examples of dithered and undithered (or very minimally dithered) uncompressed and compressed video. And there are cases where I think more dithering would have been preferable. (there's a ton of posterization in Novastorm, though the terrain often looks good, the sky looks particularly poor -I'd imagine a utility could have been used to only dither in regions where necessary)


    Quote Originally Posted by tomaitheous View Post
    I wouldn't doubt that they did manually edit some bad frames. But there's only so much you can do. And of thousands of frames, I'd hate to be the man to do it (from what you're suggesting).
    Yeah, but they could have used a software utility to optimize it instead as part of the encoding process. For compressed video I can see how dithering methods would be limited (has to cater to the block/tile nature of cinepak -and simple ordered dithering applies very well there -baer dithering using 4x4 pixel patterns)

    However, for uncompressed video at the very least (of which there are numerous examples), there's greater flexibility. The video is encoded as 16-color bitmap frames (single palette per frame -ideally each frame would use its own optimized 16-color palette), but rather than ordered dithering, a utility could be designed to convert images using things like Floyd-Seinberg (error diffusion) dithering, which still looks grainy by design, but looks much nicer as I already mentioned in my previous post. (Dragon's Lair uses it to good effect) Order dithering's simple crosshatching patter gives a lot of artifacts (rather similar to microblocking) and a posterized appearance.

    At very least for live action video it should have been used, and any cases of uncompressed video using ordered dithering. Wheither it would be preferable to optimized color reduction and simple threshhold dithering (basically undithered) with occasional use of simple checkerboard dithering for cartoon animations is another issue. I think it workded very well in Dragon's Lair, and I think could have been preferable overall to the posterization seen in most animations (including Sonic CD), though that would be up to personal taste. (cases like Sonic CD, Road Avenger, Keio Flying Squadron's intro, Road Avenger, Time Gal, Cobra Command, and such)

    In cases where the source animation is already rather low quality with simple low color graphics and simple shading, it would be preferable to probably go with threshold dithering. (Dark Wizard would be such a case)
    And cases using pixel art and tile animation in general wouldn't benefit from that either, obviously, as the graphics were drawn for the console.

    Given the comparison of the Source video of the laserdisc originals I've seen for comparison, and the full animation for Sonic CD, I'd say that good error diffusion dithering could have been beneficial for those (at least as far as my own taste goes).
    The PC version of Sonic CD's intro actually uses dithering to display it in-game with the game specific 256 color palette (apparently it sticks to that rather than loading the animation specific 256 color palette -I'm pretty sure it's encoded in an optimized 256 color indexed palette for Indeo as I've gotten win media player classic to play it in full color with windows set to 8-bit color mode)
    Though that's a bit of an apple and oranges comparison as you've got fairly good compression and a still higher color depth (albeit forced to an non-optimized one) as well as higher bitrate. (it's simple ordered dithering though -or more like a weird sub-pixel set-up in a delta pattern, like some shadow masks)


    It would be rather like nicely floyd steinberg gif animations vs ordered dithered or threshold dithered examples. (grainy but detailed appearance opposed to posterized or blocky/crosshatched) And since it's lossless compression, that's a non issue too.
    Such a utility would likely have already existed for some purpose (did photoshop support dithering and low color optimiztion at the time?), though it may have required tweaks for the specific cases used. (if there was an application which coudl take an image, select an optimum 16 color palette from 9-bit RGB and then apply floyd steinberg dithering with that palette, they'd have been set -other than setting up the utility to convert hundreds or thousands of consecutive frames)

    Again, for compressed stuff, more optimized dithering would be tough, but perhaps encoding software could have been optimized to at least do a little better than 4x4 ordered dithering. (or possibly convert the entire animation to optimized dithered uncompressed frames before compressing and then simply living with any additional artifacts -given the generally limited artifacting in cases using threshold dithering, that might have worked rather well)


    It's a limitation of the video setup. Genesis is a tile display system. It doesn't lend well for video or even stills. But at least with stills, you have more time to work the image to make it look better. The wrong colors in certain areas are conversion artifacts. Errors. Colors that couldn't be represented had to be associated to something else (preferably close). There are actually a ton of conversion errors, you just notice the bad ones. The graininess is do to the dithering process. It also helps hide errors as well as help present colors that are possibly in a 8x8 graphic block of the screen, or even colors that don't even exist in the Genesis 512 color main palette.
    In that context though it's still no more limited than a fixed 16 color indexed bitmap display (say the Atari ST) and there's the potential to do tile level optimization and effects like shadow and HL. (for digitized stills, a lot of games seem to use shadow or HL and shadow; not as good looking as 256 color images on the SNES, but a lot better than simple 16 color images too, and best cases use error diffusion dithering rather than ordered dithering)

    For animation, sure. But not for real life video. Although for some animation, dithering can be fine.
    Yeah, I think that even with ordered dithering, a case like Batman and Robbin looks nicer than had it used threshold dithering (undithered).

    Just look at Novastorm's or Stellar Fire's intro sequences to see what you end up with for live action video with just threshold dithering. (again, I think the CGI graphics in-game would have largely benefited from more dithering too -though the terrain looked pretty nice at points) Soul Star's intro seem to mostly use threshold dithering as well, or some custom optimization, but that turned out pretty well.


    Here's Novastorm and Stellar Fire for example. As I said, mainly threshold dithering with checkerboard shading being the general limit of dithering applied, and as such, any artifacting would be a fair bit more obvious than heavily ordered dithered video, but I honestly can't see much if any microblocking. ( can see it in Road Rash, though the dithering does obscure it a bit, and it's everywhere in many parts of Rebel Assault -to the point of any dithering tended to disappear into the wake of massive mircoblocking)
    So perhaps pre-convereted optimized floyd-steinberg dithered frames could have been used in conjunction with compression applied after that conversion and still look rather good. (again, cases of uncompressed video, at least ones already employing ordered dithering, certainly should have used error diffusion dithering instead -it's like a poor quick and dirty ordered dithered GIF animation vs a properly optimized one employing floyd steinberg dithering)

    http://www.youtube.com/watch?v=Qc3ECdayFL4

    http://www.youtube.com/watch?v=LD0w6towNfI


    To really get an idea of how hard it is to see artifacting, emulation set to unfiltered square pixels (unstretched on gens or using the doubleraw plugin for fusion) will demonstrate it rather clearly.

    It's interesting that i can see very obvious microblock artifacts in the PC version of Novastorm, though the video still looks far better with higher bitrate (I assume, though it was on a single disc like SCD opposed to 2 on 3DO or PSX), better compression, and 256 paletized colors. (I think the resolution might actually not be much higher as it runs in a medium sized window on a 320x200 screen)
    Edit: it seems like Novastorm's video window is almost identical, if not identical in resolution to the 256x144 Sega CD version, albeit in a different aspect ratio when viewed in the normal 320x200 4:3 aspect ratio of VGA mode 13h (with a window shaped roughly 1.48:1) compared to the Sega CD displaying in 256 pixel wide mode h32 which would result in roughly 2.07:1 in NTSC (though more like 2.22:1 in fusion without NTSC aspect enabled and on one TV I have which has vertical scan squished to show almost 240 lines but horizonal scan not adjusted -such that a 6.71 MHz dot clock has almost perfectly square pixels oddly enough matching MD mode H40), for 50 Hz PAL it would be far more stretched still, a general problem with any console outputting 256 wide resolution. (MD's 320 wide mode is a bit of a middle ground for NTSC and PAL, with pixels a bit tall in NTSC and a bit short in PAL)

    As square pixels, the 256x144 frames would be 16:9 interestingly enough.
    Last edited by kool kitty89; 06-26-2010 at 07:32 PM.
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  6. #36
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    Man, those videos remind me why I can't stand 3D from that era (cinema style). Terrible looking.

    In that context though it's still no more limited than a fixed 16 color indexed bitmap display (say the Atari ST) and there's the potential to do tile level optimization and effects like shadow and HL.
    Yeah, if you treat it as a single 16 color bitmap plane - then sure. Doing S/H requires double the vram transfer bandwidth and double the vram amount to display. It's an option sure, but not without drawbacks.

    for digitized stills, a lot of games seem to use shadow or HL and shadow; not as good looking as 256 color images on the SNES, but a lot better than simple 16 color images too, and best cases use error diffusion dithering rather than ordered dithering
    A lot? I only know of a less than handful that do.

    However, for uncompressed video at the very least (of which there are numerous examples), there's greater flexibility. The video is encoded as 16-color bitmap frames (single palette per frame -ideally each frame would use its own optimized 16-color palette), but rather than ordered dithering, a utility could be designed to convert images using things like Floyd-Seinberg (error diffusion) dithering, which still looks grainy by design, but looks much nicer as I already mentioned in my previous post.
    Just because it's uncompressed, doesn't mean it's limited to 16 colors ( 19XdGJh6Znc ). You can have uncompressed with a tilemap update too (to go along with the palette block). The easiest method, and surprisingly decent, is to divide the screen into even chucks of your total subpalette (in SegaCD case, 4 sections of the video frame).

  7. #37
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    31
    Posts
    9,725
    Rep Power
    64

    Default

    Quote Originally Posted by tomaitheous View Post
    Man, those videos remind me why I can't stand 3D from that era (cinema style). Terrible looking.
    I don't mid it so much, depending on the case. (I really like it in X-Wing and Rebel Assault to some extent -for the ships mainly, though Xwing mixes it up with a lot of pixel art and Rebel Assault does too, but in a very odd way -characters are depicted as pixel art, CGI, or digitized actors in some cases, it's rather odd)
    It's CGI characters (especially human/humanoid -or anything realistic and non cartoony/surreal) that look really off with older CGI, space Craft are probably one of the better cases (or Mechs, sci-fi tanks, vehicles, etc), some terrain is OK, and space stations or cities (especially scifi/futuristic ones) fit rather well with that modling IMO.

    Than again, I'm a rather bif fan of Mainframe Entertainment's mid/late 90s GCI animated series. (both Reboot and Beast Wars, though that's more the actual content than the animation style -which does lend itself rather well in the manner they used it IMO)

    The main point was the posterized live action shots though, compared to the 3DO version: http://www.youtube.com/watch?v=Ximd_V6I5oc


    Note I think the 3DO was the original release, though several versions could have been developed concurrently and there are considerable differences in the intro sequences. 3DO and Sega CD are similar but Sega CD adds a fair bit more clips mixed with the same overall dialog (which skips in the NTSC release it seems), while the PC/PSX releses have a totally different into using many of the same sequences added to the Sega CD version, but omit the counsel meeting thing entirely and totally change the premise of the game.
    There's gameplay differences from the Sega CD and PC/PSX versions. (not sure about 3DO, but I think it may be closer to Sega CD)


    Again, I really don't like the threshhold color conversion method used in those examples above, though it looks a little better in-game, but dithering would largely be preferable IMO.

    However, Soul Star seems to look a fair bit better in spite of being largely posterized/threshold dithered:
    http://www.youtube.com/watch?v=Iwn9RpfvEPM


    Yeah, if you treat it as a single 16 color bitmap plane - then sure. Doing S/H requires double the vram transfer bandwidth and double the vram amount to display. It's an option sure, but not without drawbacks.
    What makes it really unattractive is that it's 8 bits per pixel for only 32 colors, something the Amiga (or SNES) could manage with 5/8 the data (uncompressed). That reminds me of Chilly Willy's suggestion for an 8-bpp 1/2 res mode (using pixel accumulation somewhat like GTIA) in several hypothetical discussions on the Genesis (once in context of alternate VDP features, and other in the context of the pixel bus and an external ASIC/RAMDAC combing the pixels and outputting the video). Even a simple 8-bit RGB direct color bitmap mode like the MSX2 had could have been rather useful, especially in the context of software rendering, FMV, and the Sega CD ASIC. (allowing 3-3-2, 3-2-3, or 2-3-3 RGB to be selected would add a little flexibility too)

    A lot? I only know of a less than handful that do.
    Really, I thought a good many did. I'm positive that Toy Story did it for the intermission/cutscene images (probably HL too at least in some cases), and I'm pretty sure Earthworm Jim (and 2) did too. There are a fair number of others that look like that too, but I'd imagine for games where space was more constrained, they'd drop to 16 colors, same with SNES, a fair amout use 16 color images (not even tile optimized) but a notable amount seem to use 256 color modes (namely mode 3, but sometimes mode 4 according to BSNES -I think EWJ2 does that).
    For memory constraints, compression is an option too (or for the SNES, simply using fewer bitplanes -though that's a poor option)


    Just because it's uncompressed, doesn't mean it's limited to 16 colors ( 19XdGJh6Znc ). You can have uncompressed with a tilemap update too (to go along with the palette block). The easiest method, and surprisingly decent, is to divide the screen into even chucks of your total subpalette (in SegaCD case, 4 sections of the video frame).
    Yes, I know that, but I don't think any such games used that, and even with tile by tile optimization, that still wouldn't discount error diffusion dithering.

    So ideally, with 15/16-color tile graphics, you'd have a utility which divides the image into 8x8 tiles, selects 4 15-entry indexed palettes (with a common BG color) optimum for the tiled image, and applies those colors on a tile basis with floyd-steinberg dithering (or other error diffusion method). And possibly compress the tiles after that. (but the codec would have to specifically support the tile format and decompress to that and not a bitmap image)
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  8. #38
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    I was looking at Sherlock Holmes 2 today, for SegaCD and TurboDuo.

    Interesting things to note:

    SegaCD one uses shadow/highlight for a 128x96 pixel window.

    TurboDuo one uses 512 res pixel mode for a 240x88 window, but equiv to 120x88 window.

    Note sure about the SegaCD one, but the TCD one keeps the palette static for each scene (scene in the FMV, not per FMV instance). Rather odd choice. And although the SegaCD one has the benefit of S/H, it still has more artifacts in the video (still sub-palette limited). And even more interesting, both seem to be around the same color count per video window - roughly ~50-55 colors. And yet the SegaCD even with the same color count, still has more conversion errors (and some screens the SegaCD version has more colors by 5-10 count). Interesting.

    Any other known FMV titles on the SegaCD to use S/H?

  9. #39
    Shining Hero Joe Redifer's Avatar
    Join Date
    Dec 2005
    Location
    Denver, CO - USA
    Posts
    13,102
    Rep Power
    125

    Default

    At least the Sega CD version's audio is in sync and doesn't freeze every 5 or 6 frames.

    I don't remember any color in Sherlock Holmes other than maybe brown.

  10. #40
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    Quote Originally Posted by Joe Redifer View Post
    At least the Sega CD version's audio is in sync and doesn't freeze every 5 or 6 frames.
    Was it 1 or 2 that had those pauses? Probably 'cause the game doesn't buffer anything to ram (writes directly from CD port to vram). Huvideo and some others fixed that by absorbing any hickups with a deeper buffer/pipeline and backing off the max transfer speed (122k vs 150k).


    I don't remember any color in Sherlock Holmes other than maybe brown.
    That's cause there's a lot of wood furniture in those scenes

    Anyway, it'd be interesting to see what other SegaCD games used S/H for FMV (besides Fonzy's demos).
    Last edited by tomaitheous; 06-28-2010 at 12:14 AM.

  11. #41
    Death Bringer Raging in the Streets Black_Tiger's Avatar
    Join Date
    Oct 2006
    Location
    Vancouver
    Age
    43
    Posts
    4,946
    Rep Power
    112

    Default

    Sherlock 1 on Turbo-CD was the bad one for going out of sync as well as browniness. Sherlock 2 has some nice rich colors, but it's still like a b&w image with 4 colors painted over.

  12. #42
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    Was looking at Night Trap opening video. It's not 16 colors/single plane. The below image (the correct part) is ~45 colors, even if it doesn't look like it (which it doesn't >_>). I didn't detect any S/H colors (posterized to 3bit to find out). The SegaCD manual itself suggests using more than one layer for FMV, so that's most likely what the did here (I don't have a decent debugger SegaCD, unfortunately ).



    Even though it uses two layers (assumption), there are still conversion errors here and there. I circled them in red, so you can see them more clearly (they usually outline themselves to the tile region itself or half/part but on and edge transition making it easier to spot). Also, the frame is 168x104 and 15fps.

  13. #43
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    31
    Posts
    9,725
    Rep Power
    64

    Default

    If tile by tile optimiztion was used, that would explain the attribute clash like artifacting with the green in the upper left.

    Wouldn't using multiple BG layers complicate things and take up more memory (at least uncompressed) than only using a single layer with tile by tile palette selection? (using multiple layers might be an intersting component of a custom compression format, but for raw, uncompressed tile or bitmap video frames, that would seem almost as restricting as using shadow in terms of data size -and shadow is a bit more flexible in some other areas)

    Sewer Shark's intro does look better, but that could simply be due to the colors in the original video being drabber. (and some parts of Night Trap look better than others)

    If that was the case with the tiling and double layering, that could have made other forms of dithering tougher to implement, but I don't see why it would be totally unreasonable, even if only applied on a tile by tile basis. (rather than ordered dithering restricted to 4x4 pixel patterns, floyd-steinberg dithering could have been applied on a tile by tile basis at least, and that's assumign dropping to a single layer and/or single optimized 16-color palette and applying such dithering wouldn't still look better overall)

    This makes me wonder what other videos in general used. (for the undithered/threshold dithering examples, even with compressed stuff, I'd think that stuck to single palettes for the whole BG given the lack of tiling artifacts other than cases like Rebel Assault's massive compression artifacting)


    Also, I checked, and Wirehead is 256x224 at 10 fps, so actually slightly less compressed than Road Rash's 256x152 at 15 fps. (both have slightly variable framerates, but don't drop below those speeds, just going a bit higher in a few brief cases)


    And balck and white video was mentioned as well, and one issue with that on the Genesis is there isn't even 16 shades of gray on it (you need 12-bit RGB for that), so you're stuck with 8 shades unless you use shadow which would double the data size rather wastefully. What might be useful would be to use 3-bit planar bitmap frames and convert those to the 4-bit packed pixel tile format of the MD VDP on the fly. (I think the ASIC only handles packed pixels, so you might have to do that in software, or at least convert to a 4-bit packed bitmap frame with the CPU and use the ASIC to convert that to the VDP tile format)
    Otherwise you'd be using 4-bits per pixel for 8 shades of gray: using uncompressed planar graphics would decrease data on disc by 25%, but other than using planar graphics, actual compression algorithms could be useful in that context. (lossy or soem of the lossless compression used for tile data on carts -if that could be done fast enough for streaming video)
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  14. #44
    ding-doaw Raging in the Streets tomaitheous's Avatar
    Join Date
    Sep 2007
    Location
    Sonoran Desert
    Age
    44
    Posts
    3,981
    Rep Power
    77

    Default

    Hmm. Looks like it isn't two BG layers. I aligned the FMV window to 8x8 offset and ran it through my image converter:


    It compacted the image down to 4 subpalettes (within Genesis limits) and lossless. And no tile had more than 16 colors in it.

    If tile by tile optimiztion was used, that would explain the attribute clash like artifacting with the green in the upper left.
    That's what I mean. That's the conversion error (not enough subpalette data to represent the colors need in that tile, not that the tile itself needs more than 16 colors - that's was probably converted before this phase of the conversion process).

    Wouldn't using multiple BG layers complicate things and take up more memory (at least uncompressed) than only using a single layer with tile by tile palette selection? (using multiple layers might be an intersting component of a custom compression format, but for raw, uncompressed tile or bitmap video frames, that would seem almost as restricting as using shadow in terms of data size -and shadow is a bit more flexible in some other areas)
    Take up more VRAM memory and more vdma bandwidth, but a window that small would easily be doable. Though, the 5bit tiles would have to be split back into two 4bit tiles by one of the processors. But looking at the above pic results, it doesn't appear that way (there's not tile that has more than 16 colors in it, otherwise my app would spit out the results indicating such).

    Found something else that was interesting:




    The top image is the original, obviously. It has familiar checker board dithering. The thing with checker board dithering, is that it only takes you up 1bit per RGB element. I applied a single pixel horizontal motion blur with a value of 1 (to the native res of the pic - 256x224) of checker board pattern dither, to get back the original values that the dither was trying to represent (with a few fringes artifacts, but they very minor). That's what pic 2 is.

    The second pic had some left over horizontal lines (of the wider checker board dither). So, to figure out their color representation, I did another pass of motion blue, but this time 90 degrees instead of 0. This pic three. Pic 3 conforms exactly to a 15bit master palette. But you can still clearly see posterization. So the source image was probably 15bit color with a 256 color index (or possibly straight 15bit images).

    Now all you have to do is hack some of these old FMV games and have the 32x apply a two pass motion blur (pixel accumulator) like I did and show it on the SuperVDP instead (Excluding Night Trap because that's already has a 32x version, right?)


    And black and white video was mentioned as well, and one issue with that on the Genesis is there isn't even 16 shades of gray on it (you need 12-bit RGB for that), so you're stuck with 8 shades unless you use shadow which would double the data size rather wastefully.
    Actually, you can get a 16 'shades' of grey (and without S/H, which actually gives you less than that). You can do warm or cool expanded grey. If RGB, 8 grey scale would look like 7/7/7, 6/6/6, 5/5/5, 4/4/4, etc. But extend cool would look like 7/7/7, 6/6/7, 6/6/6, 5/5/6, 5/5/5, etc. And warm would look like 7/7/7, 7/7/6, 6/6/6, 6/6/5, 5/5/5, 5/5/4, etc. You can't mix the two (cool and warm scale) because you'd see the difference between them too much. But one or the other, works fairly decent. 'Cool' grey scale works best IMO.
    Last edited by tomaitheous; 06-28-2010 at 10:30 PM.

  15. #45
    Shining Hero Joe Redifer's Avatar
    Join Date
    Dec 2005
    Location
    Denver, CO - USA
    Posts
    13,102
    Rep Power
    125

    Default

    I always preferred the opening in the original Sega CD release of Night Trap™. It's da bomb. The openings in all other versions seems lethargic by comparison, as if the actors are super tired. Not sure why they didn't keep the "original" opening in the 32X version. It can still be the Sega Control Attack Team and have a Genesis controller.

    I made this for y'all:



    I prefer the GUI of the Sega CD version vs all other versions as well. It's easier to read and see what the hell you're doing. There is no need for the map.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •