|
No prob :) Also do early tests (with
your live psd as a texture in 'texture decal' view in full screen)
to adjust scaling of stars,
so that at the final frame res, they arent much smaller than 1
pixel (or not much bigger for that matter)
(avoiding any flickering, or avoiding having to have high sampling
to avoid flickering)
@LucEric As far as I can recall, with that env shader it was not
easy to adjust the proportions of bright vs. dark stars to make
the procedural stars not look too .. procedural :)
J
On 06/24/14 17:32, Nancy Jacobs wrote:
Thank you Jason for this awesome texturing advice. I've done
a lot in photoshop with tiles and spherical texture maps, so
this is my territory.
Some of these procedures i'll have to read over a few times
to really get completely, so I hope you don't mind if I need to
ask a couple questions about them at some point.
Thanks!
Nancy
In my experience, a textured
sphere can work pretty good,
You can tile an image (3-4 times on a sphere)
with a tilable "base star texture" (as uniform as possible)
large enough to hold enough subtle variations without
perceiving patterns (perhaps 1.5-3x the size of your final
render res),
If you are using Photoshop, from a say 1or2k rez. small-star
starfield pic,
(to make a 2-3k final pic) you can do a 'filter->offset'
by any odd amount,
and then breakup the seams to make it tilable -- super-easy
specially for stars,
you can use a speckly brush clone stamp with high opacity
(so no opacity gradient falloffs)
and a low brush step, (so 1 stamp at every ~20 pixels on
strokes for very random cloning)
So you can make a relatively 'mostly uniform ' star map
density as a base BG,
( with many-many dim (almost subpixel) stars, a a number
of mediums, and really just a couple of bright ones, all
with a bit of cloudy variations )
if there arent enough dimer ones, or to add density or
(uniformize?),
you can use a big clone stamp with that speckle-y brush, but
in additive (linear dodge) mode at varying opacity
also with that now-tilable pic, you can scale it down 50%
& tile it 4 times in half opacity (linear dodge) for
those many faint BG stars
Then, with those hubble pics, you can isolate interesting
areas, make the rest transparent,
and in 3d, add grids in key spots to add localized cloudy
nebula patterns and variations depending on what you're
after
(with RGB intensity as opacity)
If you really need 360 (up & down) with a spherical
projection,
you'll probably want to mix-in a copy of that starfield
texture for any stretching at the poles of the sphere.
I used a very speckle-y gradient (made of "fat noise") with
a white to black radial "fat noise gradient" in the center
as an alpha for the same stars texture, to project
vertically top down (x-z)
You can also blend the star textures somewhat more than 1 in
3d so that some stars can "bleed" a bit with perhaps an
additive blurred version of just those hot pixels.
That may be enough on it's own, but if you are moving around
(at light speed?)
you can also add 3D stars, Adams tips seems like an
excellent approach to that :) .. good luck! :)
Jason
On 06/23/14 17:50, Adam Sale wrote:
Do you need nebulae, etc?
If its just stars, what about using a static point
cloud with spherical / displaced randomized spheres as
shape. Randomize color and transparency per point?
This would give you the 3d field you are looking for,
then perhaps some fluids to do neb clouds, simulated
particles for comets, meteors etc..
Perhaps use the hubble images or comp some stills
together to make a bg cyclo to pull the 3d elements
together?
Adam
|
