pouët.net

__sam__ information 185 glöps

• general:
• level: user
• personal:
• first name: Samuel
• last name: Devulder

• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• @gasman, oh I see. I've reworked my dither algorithm (pseudo code)
  
  Code:-- generate all possible 4 coulor tuples tuples = [] for c1=0 to 15 do for c2=c1 to 15 do for c3=c1 to (c1<=7 ? 7 : 15) do for c4=c2 to (c2<=7 ? 7 : 15) do tuples.add([mix(c1,c3), mix(c1,c4), mix(c2,c3), mix(c2,c4)]) end end end end foreach 8x1 cell do -- dither a copy of the cell to find the best tuple best_error = infinity best_tuple = nil foreach tuple in tuples do block = copy of the 8x1 cell dither block with 4 color tuple compute(error) if best_tuple==nil or error<best_error then best_error = error best_tuple=tuple end end -- effectively dither the image dither cell with the 4 colors of the best_tuple end
  
  and it seems to work nicely. I use Otromoukhov's dither which is much better than Floyd-Steinberg.
  
  I might use these flipping attribute on the thomson TO7 :)
• isokadded on the 2017-03-26 23:17:53
• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• If I analyze correctly, not all combinations allow displaying 4 cols. And conversely not all colors are available in a 4 color cell. The selection of available colors for a cell is quite complicated. This might be a nightmare for implementing the logic in a graph tool :P
  
  Do you have any idea of the size of the palette. Is my computation (82 different colors max) correct?
• isokadded on the 2017-03-25 18:34:34
• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• @pulko 4 colors per blocks seems strange to me. I think it's more complex than that.
  
  If I understand correctly, the idea is to mix 2 color pairs: (c1,c2) + (c3,c4). There is the traditional constraint on each pair: c1 and c2 must be all bright or all dimmed (same for c3 and c4). without the constraint the theoretical palette is 5 levels per component: off+off=0% off+dim=25% dim+dim=50% dim+bright=75% and bright+bright=100% (no off+dim because it flickers too much), resulting in a 125 color palette. However the dim/bright constraint reduces his to 82. I'm working on experimental lua scripts for GraphX2 to generate the palette. You can get the whole palette I get by loading the pictures below in GraphX2
  
  It is possible to have 3 colors per block. For instance with c1/c2=black/red all dimmed and c3/c4=black/red all bright, and dimmed block is c1 c2 c2 (+ 5 others) and bright block is c3 c3 c4 (+ 5 others) we get c1+c3=black, c2+c3=25% red, c2+c4=75% red. Three colors.
  
  This is fairly complicated indeed. So to convert the pictures I simplified it by assuming that there is only one pair of colors out of the 82 palette. Notice that not all couples are valid. Some of them are to be reject. However I get some very interesting result with this simplification. Have a look at the pictures below (they all have the same 82-color palette, with only 2 colors per 8x1 block).
  
  
• isokadded on the 2017-03-25 16:37:10
• demo ZX Spectrum Compofiller - Invaders Must Die
• +1 for the (very) quick apparition of Thomson on screen. Or was it Thompson (with a typo) ? It's too quick to figure out. ;)
• rulezadded on the 2017-03-23 22:01:39
• demo SAM Coupé MusicExperiment by D.T.A. Software Studio
• I think it is the circle that is copyrighted by some ancient Greek geek :P
• isokadded on the 2017-03-22 18:52:55
• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• I'm a total newbie on Spectrum. I wonder how I can view them in 50hz. I've tried speccy 3.9 but it flickers a lot at 50hz. With 60hz it is more stable but does not show that many colors. I presume this is all related to LCD display.
  
  Mixing two images with 8x1 attribute lead to 5 intensities per components (0%=off/off, 25%=off/dimmed, 50%=dimmed/dimmed, 75%=bright/dimmed, 100%=bright/bright), so 125 cols max, but since some color combinations are forbidden on an 8x1 it's actually much less. I wonder what does the full palette looks like... Hum interesting topic :)
• isokadded on the 2017-03-22 09:15:08
• 1k ZX Spectrum Tochi by Busysoft
• Lots of variations in 1k.
• rulezadded on the 2017-03-22 07:38:59
• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• @LVD: Very interresting pics. They definitively have other colors than the basic 8+8. My guess is that these contains a palette of 16 colors out of 64 possible ones.
  
  This doesn't look like Gasman's mode, does it? The youtube video seem to use only the 8+8 colors but with color clashes of only 1 pixel tall and 8 pixel wide (eg 8x1 attribute mode).
• isokadded on the 2017-03-22 01:07:44
• demo ZX Spectrum Ultraviolet by HOOY-PROGRAM [web]
• @gasman: your new 8x1 mode seem similar to the one present on the TO7 machines (7 colors, 8x1 color clash). I've designed a conversion technique that works well with 8x1 color clashes. It is available as LUA extension to GrafX2 >>here<<
  
  Here is what the algorithm does on the same pictures you used with 7 fixed colors and 8x1 color clash
  
  
  If you wan't I can try to adapt it to the ZX.
  
  By the way big thanks to you for your JS MOD player. It inspired me to make the MOD player presented at Forever-2017 for the TO8 machine.
• rulezadded on the 2017-03-21 15:41:13
• musicdisk Thomson TO8 deMODed? by PULS [web]
• Ok, typo has been corrected.
• isokadded on the 2017-03-21 14:00:01