%OSDK%\bin\PictConv [switches] source_picture destination_fileSupported formats for the source picture is everything that FreeImage supports (BMP, PCX, PNG, TGA and TIFF among other).
- The -m switch is used to define what the target machine is:
-m0 => Oric [Default] -m1 => Atari ST
- The -f switch control the type of picture conversion that will be performed. This is machine dependent:
Oric: -f0 => to hires monochrom format [default] -f0z => to hires monochrom format (do not set bit 6) *** Not working in this version *** -f1 => precolored picture -f2 => RVB conversion -f3 => Twilight masks -f4 => RB conversion -f5 => CHAR generator (not finished) -f6 => Sam method (Img2Oric) -f7 => AIC encoding Atari ST: -f0 => Single palette format [default] -f1 => Multi palette format -f2 => Monochrome format
- The -o switch defines the final file format:
-o0 => Tape format, including a BASIC loader -o1 => Tape format, simply the picture -o2 => RAW format. No header. -o3[label name] => C source code -o4[label name] => assembler source code -o5 => Output a PC picture format -o6 => 2 bytes (dx,dy) size followed by RAW picture -o7 => palette followed by picture -o8[line:step] => BASIC source codeFor -o3 and -o4, if "label name" is not defined, the default "LabelPicture" is used.
For -o8 you can specify two values that will be taken respectively as the first line number and the step between lines.
One very useful feature is the -o5 mode which instead of saving to the Oric or Atari format, will save the result of the conversion to a pc format. This is very practical when experimenting with multi-palette or dithering options, because you do not have to try on the real machine, you can just check the result directly on your pc.
When doing Atari conversion, you typically want to use the -o7 mode, which will save the palette (or palettes) followed by the bitmap. You don't really want to try other formats because they will not work well in this version for the atari conversions (they were made for the Oric and I did not update the code yet.
- The -p switch controls the way palettes are managed:
-p0 => Generate a palette automatically [default] -p1 => Last line of the picture contains the palette -p2 => Last pixels of each line of the picture contains the palette
When using -p0, PictConv will automatically perform the palette generation, using the STe enhanced palettes, eventually reducing the color depth if more than 16 colors are used. If the color conversion is done with the -f0 format, one single palette will be generated in the final picture. If the -f1 format is used, then one palette will be generated for each scanline.
If you want to use -p1 or -p2, you need to modify your input picture to include palette data. An important point to miss, is that you can use -p1 with -f1. At first glance it would look like you would use the same single palette repeated for every scanline, which seems to be of dubious interest, but it makes sense when you will learn that the palettes you specify do not have to be complete. Only the color indexes defined in the palette will be used in the conversion process, the free indexes will be used by the color reduction routine.
When using -p1, you need to add two scanlines at the bottom of your picture. So if your picture is 320x200, you will have to make it 320x202, the two last lines will be used by PictConv to figure out which colors to use. (Of course the exported picture will be 320x200). The way it works, is that these two additional lines will contains colors that will be interpreted as either "not set" colors, or "fixed colors". The first pixel of the first additional line will be used as "not set" color, the 16 first pixels of the second additional line will be the palette itself. Each of these 16 pixels that are of the same color as the "not set" color will be ignored. Others will be interpreted as colors that have to be kept at this particular position in the palette. That's particularly useful to make sure that your background color is at index 0, and that this particular sprite is using the color 8 of the palette.
Some examples will make it easier to understand:
- The -d switch controls the dithering operation:
-d0 => No dithering (0/1) -d1 => Alternate dithering (0/0.5/1) -d2 => Ordered dithering (4x4 matrix, 8 levels) -d3 => Riemersma dithering (based on hilbert curves)
- The -a switch enable the alpha channel detection:
-a0 => no transparency [default] -a1 => encode alpha as zeroes (Only supported for monochrome and AIC pictures)
- The -t switch enable the testing mode:
-t0 => Testing disabled -t1 => Testing enabled
- The -n switch defines the number of entries per line when generating text format:
-n16 => Output 16 values each line -n40 => Output 40 values each line
- The -u switch enabled the update check:
-u0 => Do not check if files are up to date [default] -u1 => Perform a date check on the files and only update if needed
- The -v switch sets the verbosity level:
-v0 => Silent [default] -v1 => Shows information about what PictConv is doing
Here are some conversion samples, with the parameters used.
In colored mode, free size pictures are not accepted. Why ???
Try to imagine how you could scroll horizontaly a picture containing attributes changes without having huge color changes. This is not possible. Free vertical size is allowed anyway, but it's buggy right now :))
In colored mode, the converted picture contains some lines that are black and white.
These scanlines cannot be converted. A colored pictures should use Oric constraints, be in at most 8 colors, no more than 2 colors per 6 pixel bloc, and so on...
In colored mode, PictConv decided to use strange paper and ink changes.
Actually PictConv simply tries all the possible combinations of paper, ink, video inverse changes that could work to convert the picture. When it find one working combination, it moves to the next scanline.
Issue #38: Improved AIC converter
Details: It would be nice if the AIC converter supported user configurable combinations of attributes (defined in the picture as the two first pixel columns), and if it was possible to force PictConv to save the two columns with attributes so they don't have to be created by the program. And document that properly with picture examples as well!
Fixed in OSDK 1.10
Details: It seems Pictconv produces a header going from A000 to BF40 for a Hires screen, which is 1 byte too much (should stop at BF3F).
Fixed in OSDK 1.6
comments powered by Disqus