Thomas de Groot <tho### [at] degrootorg> wrote:
> Please find here the (last) version of the scene. Search for the
> differences ;-)
>
> One notable change I want to elaborate on a bit. I dumped Gilles Tran's
> MakeGrass and replaced it by Lawnmaker from Uwe Gleiss (2004).
> Unfortunately, his website from where it was downloaded does not exist
> anymore. Interestingly, the grass blades were made of parametric objects
> and I remembered the infinite time it took to render a proper scene, so
> I never used it. William Pokorny's discovery about the optimal settings
> for parametric objects, made me curious of course and, Lo! Lawnmaker
> suddenly became a serious candidate. Not stopping there, and interested
> as I am in using mesh2{} files, I decided to experiment with Meshmaker
> (thanks Ingo!). With the help of an example file by Friedrich
> Lohmueller, I made the conversion within a couple of minutes. Taking
> into account the size of the grasspatches and their grass blade density
> (which can increase parsing time quite rapidly) a typical lawn renders
> very fast, hardly slower than the rest of the scene.
>
> --
> Thomas

Hi sorry, is it too late to give suggestions? The potential of this image is big
!  My first and only remark if too late is that there would me a more immersive
feeling if silhouette of the castle had more bumpy details for slabs. That would
be a lot of work (maybe less for an HG POV displacement). So I hold my tong for
the other possible areas to work on and really appologize if it's too late, I
enjoy the meteorological effect of clouds falling down to rain a lot, and the
slopes of distant hills , and the grain on the rusty metal bar and the position
of wooden ramp to compose image into depth.

Post a reply to this message

You've definitely made some noticeable improvements on the left side.

On the right ...   it all seems to have faded into a lack of definition.
The clouds and the land ooze together into a brown cloud - I would have expected
some greater contrast, and a darker, almost black color.   Something to add
WEIGHT and authoritatively announce the arrival of a STORM, and make you _feel_
it.

You have sunlight shining on the trees, but no highlights.
An effect of the light group?

The little buildings also seem to have disappeared into the beige cloud.
Are they no longer lit as well?

It's still kicking butt, and I like both versions of the grass.

That all of it is parametric, with a conversion to mesh2{} is pretty exciting as
well.   Maybe a tiny example scene showing how all of that is implemented, along
with the needed files in a zip would prod other folks into taking a crack at
some renders of their own   ;)

Post a reply to this message

Op 15/10/2020 om 22:38 schreef Bald Eagle:
> You've definitely made some noticeable improvements on the left side.
> 
> On the right ...   it all seems to have faded into a lack of definition.
> The clouds and the land ooze together into a brown cloud - I would have expected
> some greater contrast, and a darker, almost black color.   Something to add
> WEIGHT and authoritatively announce the arrival of a STORM, and make you _feel_
> it.

Well, It may not be rain... maybe a dust storm. I understand what you 
mean but I believe I am leaving things as they are now. Maybe something 
for an update in ten years time :-)

> 
> You have sunlight shining on the trees, but no highlights.
> An effect of the light group?

No. This is something I am going to look into in my next scene which 
will be on ground level, without the castle. It probably results from 
the billboard settings and or gamma maybe.

> 
> The little buildings also seem to have disappeared into the beige cloud.
> Are they no longer lit as well?

No, they are there at the same place. What has been changed are the 
textures.

> 
> It's still kicking butt, and I like both versions of the grass.

I prefer this one .

> 
> That all of it is parametric, with a conversion to mesh2{} is pretty exciting as
> well.   Maybe a tiny example scene showing how all of that is implemented, along
> with the needed files in a zip would prod other folks into taking a crack at
> some renders of their own   ;)
> 

Excellent idea. I need to note it as todo before I forget or before RL 
kicks me over.

-- 
Thomas

Post a reply to this message

Op 15/10/2020 om 21:18 schreef Mr:
> Thomas de Groot <tho### [at] degrootorg> wrote:
>> Please find here the (last) version of the scene. Search for the
>> differences ;-)
>>
>> One notable change I want to elaborate on a bit. I dumped Gilles Tran's
>> MakeGrass and replaced it by Lawnmaker from Uwe Gleiss (2004).
>> Unfortunately, his website from where it was downloaded does not exist
>> anymore. Interestingly, the grass blades were made of parametric objects
>> and I remembered the infinite time it took to render a proper scene, so
>> I never used it. William Pokorny's discovery about the optimal settings
>> for parametric objects, made me curious of course and, Lo! Lawnmaker
>> suddenly became a serious candidate. Not stopping there, and interested
>> as I am in using mesh2{} files, I decided to experiment with Meshmaker
>> (thanks Ingo!). With the help of an example file by Friedrich
>> Lohmueller, I made the conversion within a couple of minutes. Taking
>> into account the size of the grasspatches and their grass blade density
>> (which can increase parsing time quite rapidly) a typical lawn renders
>> very fast, hardly slower than the rest of the scene.
>>
>> --
>> Thomas
> 
> Hi sorry, is it too late to give suggestions? The potential of this image is big
> !  My first and only remark if too late is that there would me a more immersive
> feeling if silhouette of the castle had more bumpy details for slabs. That would
> be a lot of work (maybe less for an HG POV displacement). So I hold my tong for
> the other possible areas to work on and really appologize if it's too late, I
> enjoy the meteorological effect of clouds falling down to rain a lot, and the
> slopes of distant hills , and the grain on the rusty metal bar and the position
> of wooden ramp to compose image into depth.
> 
> 

Never too late for suggestions; if I am going to do something with them 
is another matter. ;-)

For those parts of the castle which show a definite shadow in the Sun 
(the tower) I used the displacement feature in Poseray. This is tricky 
business because you can easily overdo it and the effect becomes worse. 
So, the roof tiles and the stones in the wall were subtly displaced. The 
shadowy wall were not displaced though. Overall, I applied a proximity 
pattern to make the wall texture more interesting.

There are always more features to work upon in such a scene. I have a 
couple of items myself. However, before I get bored I prefer to stop now 
and maybe take up the work later (or never). Perfection is not my cup of 
tea. :-)

-- 
Thomas

Post a reply to this message

Op 15/10/2020 om 09:26 schreef Kenneth:
> Thomas de Groot <tho### [at] degrootorg> wrote:
>> Please find here the (last) version of the scene. Search for the
>> differences ;-)
>>
> 
> Even more beautiful than your original render.
> 
> I see that the trees have been fattened-up a bit; and one of the rusty metal
> fence-bars has been broken and bent (a nice touch.) It also looks like the
> castle or church has been thrown into more shadow (or maybe with less fill
> light), which nicely matches the late afternoon Sun. The stones under the fence
> seem to have changed too.

Different changes here: besides fattening up of the trees, I threw in a 
second one and both are randomly selected. the castle walls have an 
added proximity pattern to the texture (the Edouard Poor variant). It 
makes them less "clean". New stones under the fence indeed, and I kicked 
the metal bars into shape in some places, breaking it at one place (sorry).

> 
> I do like the detailed grass, but my personal preference is for the previous
> 'mossy' appearance; it just seems to go better with the overall 'look' of the
> scene-- maybe because, with its 'lack of detail', it offsets the more-detailed
> valley and trees. And it more clearly shows the shadow of the out-of-frame tree
> that is cast across the lawn.
> 
> I like *both* versions of the fence stones; I think I prefer the previous
> stones' textures, but the new stones' shapes. A difficult decision!

Ha! Here the supremely powerful might of the author comes into play. I 
shall leave it as-is.

> 
> It's easy to be an amateur art critic, ha :-P

Sure! Which is nonetheless fully appreciated as it often points directly 
to sore places which need treatment.

> 
> You have used so *many* interesting code tools for this scene, tools that I have
> not even started experimenting with yet. Thanks for the detailed descriptions of
> your techniques.
> 

You are most welcome. I need to do something about all this. I have made 
a small start into preparing all the stuff for more general use. I met a 
lot of loose ends and inconsistencies already so, that is going to take 
some time...


-- 
Thomas

Post a reply to this message

Thomas de Groot <tho### [at] degrootorg> wrote:
> Op 15/10/2020 om 22:38 schreef Bald Eagle:
> >
> > You have sunlight shining on the trees, but no highlights.
> > An effect of the light group?
>
> No. This is something I am going to look into in my next scene which
> will be on ground level, without the castle. It probably results from
> the billboard settings and or gamma maybe.
>

An idea off the top of my head: Instead of a flat billboard, you could make a
simple low-rez height_field for the tree (including the trunk too, possibly).
Perhaps 2 or 3 different ones, for variation. (Pre-#declare each HF so that it
doesn't take up 'extra' memory when used repeatedly.) Then the trees would show
some highlights, using phong or whatever.

Since your current billboard scheme has the trees always facing the camera, the
same scheme would be used for the HF 'shells'-- the 'rear' or underside of the
HFs would never be seen.

Post a reply to this message

"Kenneth" <kdw### [at] gmailcom> wrote:

> An idea off the top of my head: Instead of a flat billboard, you could make a
> simple low-rez height_field for the tree (including the trunk too, possibly).

Hmmm.
If a heightfield is just a displacement based on the grayscale value of the
image pixels, then you could also make an isosurface of it, and abs() should
mirror the
displacement in both directions.
Then I think there's the isosurface approximation macros that could make a
mesh{} of _that_, and then you could possibly make a whole 3D tree from just the
image.
Alternatively, I guess you could just intersect{} it with a box and texture it
with the image_map and see how that turns out...

Just thinking out loud..

Post a reply to this message

Le 2020-10-16 à 02:42, Thomas de Groot a écrit :
> Op 15/10/2020 om 21:18 schreef Mr:
>> Thomas de Groot <tho### [at] degrootorg> wrote:
>>> Please find here the (last) version of the scene. Search for the
>>> differences ;-)
>>>
>>> One notable change I want to elaborate on a bit. I dumped Gilles Tran's
>>> MakeGrass and replaced it by Lawnmaker from Uwe Gleiss (2004).
>>> Unfortunately, his website from where it was downloaded does not exist
>>> anymore. Interestingly, the grass blades were made of parametric objects
>>> and I remembered the infinite time it took to render a proper scene, so
>>> I never used it. William Pokorny's discovery about the optimal settings
>>> for parametric objects, made me curious of course and, Lo! Lawnmaker
>>> suddenly became a serious candidate. Not stopping there, and interested
>>> as I am in using mesh2{} files, I decided to experiment with Meshmaker
>>> (thanks Ingo!). With the help of an example file by Friedrich
>>> Lohmueller, I made the conversion within a couple of minutes. Taking
>>> into account the size of the grasspatches and their grass blade density
>>> (which can increase parsing time quite rapidly) a typical lawn renders
>>> very fast, hardly slower than the rest of the scene.
>>>
>>> -- 
>>> Thomas
>>
>> Hi sorry, is it too late to give suggestions? The potential of this 
>> image is big
>> !  My first and only remark if too late is that there would me a more 
>> immersive
>> feeling if silhouette of the castle had more bumpy details for slabs. 
>> That would
>> be a lot of work (maybe less for an HG POV displacement). So I hold my 
>> tong for
>> the other possible areas to work on and really appologize if it's too 
>> late, I
>> enjoy the meteorological effect of clouds falling down to rain a lot, 
>> and the
>> slopes of distant hills , and the grain on the rusty metal bar and the 
>> position
>> of wooden ramp to compose image into depth.
>>
>>
> 
> Never too late for suggestions; if I am going to do something with them 
> is another matter. ;-)
> 

Maybe not for this one, but for your next master piece 😉

Post a reply to this message

Op 16/10/2020 om 22:42 schreef Bald Eagle:
> "Kenneth" <kdw### [at] gmailcom> wrote:
> 
>> An idea off the top of my head: Instead of a flat billboard, you could make a
>> simple low-rez height_field for the tree (including the trunk too, possibly).
> 
> Hmmm.
> If a heightfield is just a displacement based on the grayscale value of the
> image pixels, then you could also make an isosurface of it, and abs() should
> mirror the
> displacement in both directions.
> Then I think there's the isosurface approximation macros that could make a
> mesh{} of _that_, and then you could possibly make a whole 3D tree from just the
> image.
> Alternatively, I guess you could just intersect{} it with a box and texture it
> with the image_map and see how that turns out...
> 
> Just thinking out loud..
> 
> 

Yes, all good suggestions needing to be tested... I have a bit too many 
things on my daily plate to be able to chew them properly at this 
moment. Starting a /new/ ToDo list (as if that would help).

-- 
Thomas

Post a reply to this message

"Bald Eagle" <cre### [at] netscapenet> wrote:
> "Kenneth" <kdw### [at] gmailcom> wrote:
>
> > An idea off the top of my head: Instead of a flat billboard, you could make a
> > simple low-rez height_field for the tree (including the trunk too, possibly).
>
> Hmmm.
> If a heightfield is just a displacement based on the grayscale value of the
> image pixels, then you could also make an isosurface of it, and abs() should
> mirror the
> displacement in both directions.
> Then I think there's the isosurface approximation macros that could make a
> mesh{} of _that_, and then you could possibly make a whole 3D tree from just the
> image.
> Alternatively, I guess you could just intersect{} it with a box and texture it
> with the image_map and see how that turns out...
>

I was actually thinking that a sort of dome-shaped, low-rez crinkly HF (with
'smooth', and 'water_level' to clip off the ground plane), could then be rotated
-90*x, then aligned to the camera view like Thomas did with his billboards. With
the tree texture applied, that would make a quasi-3D tree shape, at least for
the 'bulk of tree leaves' if not the trunk, and it might pick up some
semi-realistic lighting highlights if phong is used. It wouldn't matter that the
HF is just a shell; its underside ('rear') would never be seen, since the HF
would always face the camera.

Post a reply to this message