Prerelease talk:Super Mario World (SNES)/Background Graphics and Tilemaps

Improvised Tiles for Early SCR files

assets involved

• \NEWS_11\hino\z-mario-4
• kinoko-TEST.scr
• v-ram-map-1.col
• CLR.CGX
• X-I-P-NES.cgx
• Rearranged tiles based on X-I-P-NES.cgx
• https://tcrf.net/images/e/ea/SMWProto_Map.jpg
• CGX Viewer
• YY-CHR
• Lunar Magic

The key to this process was letting the screen file tell me what tiles it needed. This methodology is what makes this file (almost!) authentic.

1) First, I made a copy of CLR.cgx, and titled it 'KINOKOTESTNOTASSET.CGX'. I then filled out this file with a hex graphic for each tile ID. I loaded this in CGX Viewer to get an idea of the tilemap and its attribute info of 'kinoko-TEST.scr'. This let the tilemap show me metadata needed to reconstruct the map, such as which palette was used for which tile, whether the tile was mirrored, and the location the tile was drawn.

2) Then I took that hex tilemap and compared it against the early scan image of a similar island I linked above. Then I opened'X-I-P-NES.cgx' and used it as a place to pull map tiles from for reconsting the map. Thanks to @BrunoValads brilliant suggestion, I also used Lunar Magic to verify how map tiles were used in the final game, also as a reference.

3) The green hills were the first place to start, as they had a distinct pattern I could reconstruct. Then I moved on to the detailed grass and the sloped, angled hills. This gave me a great starting reference for the rest of my workflow.

3) Some cliffs had to be custom made to work with the map. The final game is pretty clever with reusing cliff tiles efficiently, including mirroring and palette swapping. 'kinoko-TEST.scr' uses similar looking tiles, but instead of mirroring them it appears to have custom tiles drawn in their place, achieving the same effect. Some tiles are wasteful, such as repeating single color tiles multiple times, or using a single tile for a small corner detail. It's entirely possible I mistook some information due to the quality of the scan.

I also had to create graphics to fit the ponds and detailed grass in the scan. These are not found in the leaked assets, and could not be used. I also did not try and make sure the palettes matched what was shown in the scan--I just wanted insert graphic information to complete the .scr file.

I think I did as much with kinoko.scr and kinoko-test.scr as I can. there's a single tile I can't find any indication what it should be, so I left it as a large X tile. Here they are viewed with x-color-e.col and v-ram-map-2.col.

There are two small path tiles that have a blue color for grass instead of green. I tried all the palette swatches available, but none would allow the green color needed for the grass. This seems to be an attribute problem with the SCR files themselves, but it is possible I made an error. If anyone wants to verify, please do!