Gaseous Giganticus

[ILikeSaturn]

I compiled Gaseous Giganticus with MinGW:

gaseous-giganticus.exe

(568.81 KiB) Downloaded 516 times

I analyzed the exe and it requires no extra dlls
It's also compatible with 32 bit systems
Unfortunately, it was designed for powerful PCs. But, it will try to adjust the memory usage when generating a texture. If you choose a height over 2048 pixels for the -E option, it might crash.

It uses the same seed for texture generation, so expect the same patterns. I will contact the developer soon to add a custom seed option.

The input file must be in png. It's recommended that the png is a gradient.
Instructions (options are case sensitive):

Code: Select all

usage: gaseous-giganticus [-b bands] [-i inputfile] [-o outputfile]
       [-w w-offset] [-h] [-n] [-v velocity factor] [-B band-vel-factor]

Options:
   -a, --pole-attenuation: attenuate band velocity near poles.  If
                 attenuation is zero, band velocity at poles will be
                 relatively high due to short distance around band.
                 Default is 0.5. Min is 0.0, max is 1.0.
   -b, --bands : Number of counter rotating bands.  Default is 6.0
   -B, --band-vel-factor: Multiply band velocity by this number when
                   computing velocity field.  Default is 2.9
   -c, --count : Number of iterations to run the simulation.
                 Default is 1000
   -C, --cloudmode: modulate image output by to produce clouds
   -d, --dump-velocity-field : dump velocity field data to specified file
                               (see -r option, below)
   -D, --faderate : Rate at which particles fade, default = 0.01
   --dump-flowmap : dump velocity field data to a series of
                    six RGB png files encoding x, y in R, G channels
   -e, --band-speed-power: Band speed is modulated by
        cos(K*latitude) ^ band-speed-power, where band-speed-power is an
        odd integer. By default, 1, higher values make the fast moving part
        of the bands narrow and the parts between wider.
   -E, --equirectangular image-height. Output an equirectangular image in
         addition to the usual cubemap images.  The image height must be
         an integer power of 2.
   -f, --fbm-falloff: Use specified falloff for FBM noise.  Default is 0.5
   -F, --vfdim: Set size of velocity field.  Default:2048. Min: 16. Max: 2048
   -g, --gain, 2nd and later octaves are multiplied by pow(fbm-falloff, (octave-1)*gain)
   -h, --hot-pink: Gradually fade pixels to hot pink.  This will allow
                   divergences in the velocity field to be clearly seen,
                   as pixels that contain no particles wil not be painted
                   and will become hot pink.
   -i, --input : Input image filename.  Must be RGB or RGBA png file.
   -I, --image-save-period: Interval of simulation iterations after which
         to output images.  Default is every 25 iterations
   -k, --cubemap: input 6 RGB or RGBA png cubemap images to initialize particle color
   -l, --large-pixels: particles are 3x3 pixels instead of 1 pixel.  Allows
                       for fewer particles and thus faster rendering, but this
                       WILL leave visible artifacts at cubemap face boundaries.
   -L, --noise-levels: Number of fractal noise levels to consider.  Default 4, max 7
   -m, --speed-multiplier:  band_speed_factor and velocity_factor are
         by this number.  It is a single option to affect both by a
         multiplier which is a bit easier than setting an absolute value
         for them individually.
   -n, --no-fade:  Do not fade the image at all, divergences will be hidden
   -N, --sequence-number:  Do not overwrite output images, instead embed a
                           sequence number in the filenames.
   -o, --output : Output image filename template.
               Example: 'out-' will produces 6 output files
               out-0.png, out-1.png, ..., out-5.png
   -O, --opacity: Specify minimum opacity of particles between 0 and 1.
                  default is 0.2
   -p, --particles: Use specified number of particles. Default is 8000000.
   -P, --plainmap  Do not use sinusoidal image mapping, instead repeat image
                   on six sides of a cubemap.
   -r, --restore-velocity-field: Instead of computing the velocity field from
                                 scratch, import a previously saved velocity
                                 field (see -d option, above)
   -R, --random: Random values are used for bands, band-vel-factor, velocity-factor
                 noise-scale, and w-offset.  -S and -V options are also set.
   -s, --stripe: Begin with stripes from a vertical strip of input image
   -S, --sinusoidal: Use sinusoidal projection for input image
                 Note: sinusoidal is the default projection.
                 Note: --stripe and --sinusoidal are mutually exclusive
   -t, --threads: Use the specified number of CPU threads up to the
                   number of online CPUs
   -T, --thread-iterations: Number of iterations of thread movement particles
         before threads join. Default is 1. Higher values will reduce the number
         of times threads are created and destroyed.
   --trap-nans: trap divide by zero, overflow and invalid floating point
                exceptions
   -v, --velocity-factor: Multiply velocity field by this number when
                   moving particles.  Default is 1200.0
   -V, --vertical-bands:  Make bands rotate around Y axis instead of X
                          Also affects --stripe option.
   --vortex-band-threshold: controls distribution of vortices (see man page)
   --vortex-size: radius of vortices as a fraction of planet radius. Default 0.04
   --vortex-size-variance: Range of vortex sizes, default is plus or minus 0.02
   -w, --w-offset: w dimension offset in 4D open simplex noise field
                   Use -w to avoid (or obtain) repetitive results.
   -W, --wstep: w coordinate of noise field is incremented by specified
                amount periodically and velocity field is recalculated
   -x, --vortices: how many artificial circular vortices to add into the v-field
   -z, --noise-scale: default is 2.600000

Note: As noise-scale increases, velocity-factor should generally decrease.

Examples:

   gaseous-giganticus --noise-scale 2.5 --velocity-factor 1300 --bands 10\
        -i input_image.png -o output_image
   gaseous-giganticus -R -i input_image.png -o output_image


If you don't want to read the instructions, you can use this simple command:

Code: Select all

gaseous-giganticus -i [input PNG file] -o [output name] -E 1024 -c 400

Some examples:

Tips:
If you want to reduce gaps when generating a texture, use the -l option. Only use it if you're generating a 1k texture.
If you want to generate an equirectangular texture, use the -E option. Just specify the height of the texture map.

[Tegmine]

You can adjust number of bands, the scale and the velocity of those to achieve different looks. It isn't exactly "the same seed." You do have a measure of control over it.

BTW, your gas giant textures look very nice, but I especially like the third one.

-M-

[smcameron]

> I will contact the developer soon to add a custom seed option.

Hello. I am the author of gaseous-giganticus. You have not contacted me about this proposed custom seed option (it's been more than 30 days since your post stating that you would contact me.)

In any case, the -w option, "w dimension offset in 4D open simplex noise field" can serve this purpose. There are enough other variables (e.g. --bands, or the input image) that it's typically not necessary to worry about this, the generated textures are different enough anyway. You'd have to work pretty hard to get gaseous-giganticus to generate textures that were too similar.

> Unfortunately, it was designed for powerful PCs.

It's not really the case that it's "designed" for powerful PCs, it's just that there's a *lot* of math required and there's just no avoiding it. Math takes time. If you want results quickly, you need a fast computer. A slow computer will still work though, it will just take longer to get the same results as a faster computer.

[John Van Vliet]

even with 8 cpu cores gaseous-giganticus will take a few min. to run , but that is ok .

thanks smcameron for writing the program

[Tegmine]

I concur. I have found it takes no more than a few minutes on my clunky old beast of a laptop! It's a matter of knowing when to pull the plug on the process.

-M-

[ILikeSaturn]

Recompiled it from the latest commit

gaseous-giganticus.zip

(263.01 KiB) Downloaded 309 times

Usage:

Spoiler

Code: Select all

usage: gaseous-giganticus [-b bands] [-i inputfile] [-o outputfile]
       [-w w-offset] [-h] [-n] [-v velocity factor] [-B band-vel-factor]

Options:
   -a, --pole-attenuation: attenuate band velocity near poles.  If
                 attenuation is zero, band velocity at poles will be
                 relatively high due to short distance around band.
                 Default is 0.5. Min is 0.0, max is 1.0.
   -b, --bands : Number of counter rotating bands.  Default is 6.0
   -B, --band-vel-factor: Multiply band velocity by this number when
                   computing velocity field.  Default is 2.9
   -c, --count : Number of iterations to run the simulation.
                 Default is 1000
   -d, --dump-velocity-field : dump velocity field data to specified file
                               (see -r option, below)
   -D, --faderate : Rate at which particles fade, default = 0.01
   --dump-flowmap : dump velocity field data to a series of
                    six RGB png files encoding x, y in R, G channels
   -e, --band-speed-power: Band speed is modulated by
        cos(K*latitude) ^ band-speed-power, where band-speed-power is an
        odd integer. By default, 1, higher values make the fast moving part
        of the bands narrow and the parts between wider.
   -E, --equirectangular image-height. Output an equirectangular image in
         addition to the usual cubemap images.  The image height must be
         an integer power of 2.
   -f, --fbm-falloff: Use specified falloff for FBM noise.  Default is 0.5
   -F, --vfdim: Set size of velocity field.  Default:2048. Min: 16. Max: 2048
   -g, --gain, 2nd and later octaves are multiplied by pow(fbm-falloff, (octave-1)*gain)
   -h, --hot-pink: Gradually fade pixels to hot pink.  This will allow
                   divergences in the velocity field to be clearly seen,
                   as pixels that contain no particles wil not be painted
                   and will become hot pink.
   -i, --input : Input image filename.  Must be RGB or RGBA png file.
   -I, --image-save-period: Interval of simulation iterations after which
         to output images.  Default is every 25 iterations
   -k, --cubemap: input 6 RGB or RGBA png cubemap images to initialize particle color
   -K, --cache-aware fraction:  initialize the first fraction of particles in a cache
         aware way (faster) and remaining particles with random locations.
         E.g. -K 0.5 will initialize 50% of particles in a cache aware way and
         50% randomly.
   -l, --large-pixels: particles are 3x3 pixels instead of 1 pixel.  Allows
                       for fewer particles and thus faster rendering, but this
                       WILL leave visible artifacts at cubemap face boundaries.
   -L, --noise-levels: Number of fractal noise levels to consider.  Default 4, max 7
   -m, --speed-multiplier:  band_speed_factor and velocity_factor are
         by this number.  It is a single option to affect both by a
         multiplier which is a bit easier than setting an absolute value
         for them individually.
   -n, --no-fade:  Do not fade the image at all, divergences will be hidden
   -N, --sequence-number:  Do not overwrite output images, instead embed a
                           sequence number in the filenames.
   -o, --output : Output image filename template.
               Example: 'out-' will produces 6 output files
               out-0.png, out-1.png, ..., out-5.png
   -O, --opacity: Specify minimum opacity of particles between 0 and 1.
                  default is 0.2
   -p, --particles: Use specified number of particles. Default is 8000000.
   -P, --plainmap  Do not use sinusoidal image mapping, instead repeat image
                   on six sides of a cubemap.
   -r, --restore-velocity-field: Instead of computing the velocity field from
                                 scratch, import a previously saved velocity
                                 field (see -d option, above)
   -R, --random: Random values are used for bands, band-vel-factor, velocity-factor
                 noise-scale, and w-offset.  -S and -V options are also set.
   -s, --stripe: Begin with stripes from a vertical strip of input image
   -S, --sinusoidal: Use sinusoidal projection for input image
                 Note: sinusoidal is the default projection.
                 Note: --stripe and --sinusoidal are mutually exclusive
   -t, --threads: Use the specified number of CPU threads up to the
                   number of online CPUs
   -T, --thread-iterations: Number of iterations of thread movement particles
         before threads join. Default is 1. Higher values will reduce the number
         of times threads are created and destroyed.
   --trap-nans: trap divide by zero, overflow and invalid floating point
                exceptions
   -v, --velocity-factor: Multiply velocity field by this number when
                   moving particles.  Default is 1200.0
   -V, --vertical-bands:  Make bands rotate around Y axis instead of X
                          Also affects --stripe option.
   --vortex-band-threshold: controls distribution of vortices (see man page)
   --vortex-size: radius of vortices as a fraction of planet radius. Default 0.04
   --vortex-size-variance: Range of vortex sizes, default is plus or minus 0.02
   -w, --w-offset: w dimension offset in 4D open simplex noise field
                   Use -w to avoid (or obtain) repetitive results.
   -W, --wstep: w coordinate of noise field is incremented by specified
                amount periodically and velocity field is recalculated
   -x, --vortices: how many artificial circular vortices to add into the v-field
   -z, --noise-scale: default is 2.600000

Note: As noise-scale increases, velocity-factor should generally decrease.

Examples:

   ./gaseous-giganticus --noise-scale 2.5 --velocity-factor 1300 --bands 10\
        -i input_image.png -o output_image
   ./gaseous-giganticus -R -i input_image.png -o output_image

Hints for speeding things up (part of the help message)

Spoiler

Code: Select all

Hints for speeding things up:

0. If your system does CPU frequency scaling, make sure whatever governs this is
   in a 'performance' mode rather than a 'power saving' mode. (On linux, see cpufreq-set)
1. First choose bands, velocity-factor, and noise-scale with a small number of
   particles, say, -p 50000. This will allow you to quickly iterate these values
   to get the general shape of the turbulence and overall look of the output the
   way you want it, and to determine roughly how many iterations are needed.
2. Once you establish roughly how many iterations you need (200-250 is usually good)
   use the -c and -I options.  PNG encoding takes significant time, so doing it
   less frequently speeds things up considerably.
3. Once you have the general shape of the turbulence, use the -d option to save
   the velocity field to a file, and in subsequent runs, use -r to load it from
   this file rather than calculating it from scratch each time.
4. Next iterate on your input image colors.  For this, use -K 1.0 with 2500000 to
   3000000 particles or so.
5. Once satisfied with your input image, run without the -K option and with the default
   of 8 million particles for better coverage and smoother output.