Author Topic: Apoco77 House  (Read 44568 times)

2016-09-14, 01:25:29
Reply #45

zuliban

  • Active Users
  • **
  • Posts: 56
    • View Profile
Thank you Ondra,  just take your time and do your pony magic ;). im sure you will come eventually with something really good.

2016-09-14, 03:17:43
Reply #46

agentdark45

  • Active Users
  • **
  • Posts: 577
    • View Profile
this is a personal one but the ''filmic'' tone mapping for me is a wrong step for Corona, Fstorm got a correct tone mapping almost identical to a camera what is lacking now is the camera responses like what octane have.

Just had a quick look at the F-Storm gallery and I am completely amazed by how photographic the renders look! Whatever tone mapping implementation they are using we desperately need it in Corona!
Vray who?

2016-09-14, 12:20:11
Reply #47

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
sorry zuliban for the cross posting...

agentdark45, in my opinion, you can obtain the same rendering and tone mapping in Corona
Look at my comparison.

LINK

Bye

2016-09-14, 14:16:56
Reply #48

Juraj

  • Active Users
  • **
  • Posts: 4765
    • View Profile
    • studio website
The tonemapping in F-storm is like longer shoulder filmic. ( Reinhard has really long shoulder that absolutely flattens highlight to midpoints, hence why it is so dull and nasty, and VFB+ Filmic has too short shoulder, so it will only dull the very highlights at the right edge of histogram, this might not look very natural either for scenes with very strong dynamic range ).

I haven't yet tested how the filmic version in Corona version compares, but apparently it's bit nicer also. I still want to see ACES filmic curve which should be good film replica, bit more tuned that Habble's version.
Please follow my new Instagram for latest projects, tips&tricks, short video tutorials and free models
Behance  Probably best updated portfolio of my work
lysfaere.com Please check the new stuff!

2016-09-14, 15:32:22
Reply #49

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12783
  • Marcin
    • View Profile
another thing i do not like was the removal of the ''low thread priority'' this is why i used 1.3 too ,this scene is big and ram intensive so it used more than my system had and 1.4 parsing times where even 1 hour vs 1.3 4min.
This should never happen, and we are looking for users with exactly this problem. Is it ok if someone from our team contacts you to look further into this issue?

what i do not like is the removal of the bucket engine.... i rendered all these images using 1.3 bucket  it was also better for preview since it has somekind of AA while the progressive you need to wait more to have a decent AA to see the general shape better maybe im wrong since i have used buckets for so much time.
I think this is a misconception. Buckets worked on exactly the same principle as progressive rendering, but with slightly different settings, and with very basic implementation of adaptivity. The main differences were:
-setting up desired AA in advance (this would however produce exactly the same results as progressive rendering after a specific number of passes)
-quick preview of small region vs slower refinement of the whole screen (this can be however achieved using region rendering now)
-adaptivity (there is much better adaptivity implemented in progressive rendering now)
So there are virtually no advantages of bucket rendering over progressive. I am really curious what would happen if you would switch to 1.4 now (or even better 1.5 dailies) and re-render some of your awesome shots. :)
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2016-09-14, 16:07:47
Reply #50

Nekrobul

  • Primary Certified Instructor
  • Active Users
  • ***
  • Posts: 1028
    • View Profile
another thing i do not like was the removal of the ''low thread priority'' this is why i used 1.3 too ,this scene is big and ram intensive so it used more than my system had and 1.4 parsing times where even 1 hour vs 1.3 4min.
This should never happen, and we are looking for users with exactly this problem. Is it ok if someone from our team contacts you to look further into this issue?

what i do not like is the removal of the bucket engine.... i rendered all these images using 1.3 bucket  it was also better for preview since it has somekind of AA while the progressive you need to wait more to have a decent AA to see the general shape better maybe im wrong since i have used buckets for so much time.
I think this is a misconception. Buckets worked on exactly the same principle as progressive rendering, but with slightly different settings, and with very basic implementation of adaptivity. The main differences were:
-setting up desired AA in advance (this would however produce exactly the same results as progressive rendering after a specific number of passes)
-quick preview of small region vs slower refinement of the whole screen (this can be however achieved using region rendering now)
-adaptivity (there is much better adaptivity implemented in progressive rendering now)
So there are virtually no advantages of bucket rendering over progressive. I am really curious what would happen if you would switch to 1.4 now (or even better 1.5 dailies) and re-render some of your awesome shots. :)


We with Ondra have made a lot of testing searching for the cause of this problem. And this definetly was not LTP switch (wich btw remains in the setting textfile).

There is something other going on, but to fing that out there is need to test all of the 1.3 to 1.4 daily builds and read carefully changelogs.

I just don't have any of the spare time to continue testing (for now)

There is hidden massive bug wich affects CPU work not only memory usage but it is tricky and pops up only when memory runs out.
---------------------------------------------------------------
https://www.blackbellstudio.com/
https://www.behance.net/blackbell3d
CEO at "Blackbell Studio"

2016-09-14, 16:16:58
Reply #51

GabaCGStudio

  • Guest
need to test all of the 1.3 to 1.4 daily builds and read carefully changelogs.
There is hidden massive bug wich affects CPU work not only memory usage but it is tricky and pops up only when memory runs out.

when the embree was changed to new embree code (in mid of 1.4 DB) this problem was bigger!!

2016-09-14, 16:26:14
Reply #52

Nekrobul

  • Primary Certified Instructor
  • Active Users
  • ***
  • Posts: 1028
    • View Profile
need to test all of the 1.3 to 1.4 daily builds and read carefully changelogs.
There is hidden massive bug wich affects CPU work not only memory usage but it is tricky and pops up only when memory runs out.

when the embree was changed to new embree code (in mid of 1.4 DB) this problem was bigger!!

This is definetly not new embree, embree was the first thing we tested.

Something before it. 3-4-5 DB before.

Worst case scenario is when it is several things causing it wich have apeard in different time and are giving minor preformance drop. It his case it will be very hard to indentify it.
« Last Edit: 2016-09-14, 16:30:22 by Nekrobul »
---------------------------------------------------------------
https://www.blackbellstudio.com/
https://www.behance.net/blackbell3d
CEO at "Blackbell Studio"

2016-09-14, 16:42:51
Reply #53

GabaCGStudio

  • Guest
This is definetly not new embree, embree was the first thing we tested.

tested new embree with dozens of scattered proxies? with or without coronascatter or others?

or BSP tree bug? BSP depth??
« Last Edit: 2016-09-14, 16:47:47 by GabaCGStudio »

2016-09-14, 16:52:02
Reply #54

Nekrobul

  • Primary Certified Instructor
  • Active Users
  • ***
  • Posts: 1028
    • View Profile
This is definetly not new embree, embree was the first thing we tested.

tested new embree with dozens of scattered proxies? with or without coronascatter or others?

In the same scene with insane amount of scattered objekts With massive overload of memmory (24 GB usage with 16 GB RAM pool)
---------------------------------------------------------------
https://www.blackbellstudio.com/
https://www.behance.net/blackbell3d
CEO at "Blackbell Studio"

2016-09-14, 16:53:58
Reply #55

agentdark45

  • Active Users
  • **
  • Posts: 577
    • View Profile
sorry zuliban for the cross posting...

agentdark45, in my opinion, you can obtain the same rendering and tone mapping in Corona
Look at my comparison.

LINK

Bye

I don't know man, I see some pretty drastic differences between the comparisons (glow/glare and textures aside). Colour richness, highlights, shading, contrast - all seem to be lacking in the Corona examples compared to the Fstorm render which looks quite a bit more photographic. I'm sure you could get fairly close to the Fstorm renders with some additional post processing, but it would be nice to have an "out of the box" no major tweaking needed solution.

Not that I'm dinging Corona or the dev team in anyway - it's still the best render engine I've ever used. I would just like to see more development of the tone mapping side of things as it's such a game changer to the overall look of our renders. I feel like emulating how film/SLR's work in the real world is one of the most overlooked aspects of rendering engines.

I don't think this is too off topic as Zuliban mentioned it.

Interesting stuff re the ACES Filmic mapping Juraj!
« Last Edit: 2016-09-14, 17:01:12 by agentdark45 »
Vray who?

2016-09-14, 16:55:41
Reply #56

GabaCGStudio

  • Guest
In the same scene with insane amount of scattered objekts With massive overload of memmory (24 GB usage with 16 GB RAM pool)

so ... we can't set BSP depth manually ... probably need to tweak BSP depth!?

2016-09-14, 17:07:05
Reply #57

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
we will try to squeeze in the LUT mapping into 1.5, stay tuned ;)
Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2016-09-14, 17:43:48
Reply #58

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
In my test, the goal was to test the rendering time, not the tone mapping.
In my scene, the main "errors"  was in the material setting.
For example, the red plastic, in my scene, is too dark. The same for the floor, wood etc...
Also, when I did the test, Corona 1.4 hasn't bloom and Filmic.
In the Fstorm scene, I think that there was a yellow light outside ( I don't remember exactly)

I'm sure that we are able to make the Corona render very close to FStorm.
Feel free to open my scene and make it 100% similar FStorm. Now I have no time and I'm not interested .

2016-09-14, 18:08:02
Reply #59

zuliban

  • Active Users
  • **
  • Posts: 56
    • View Profile
i knew somekind of renderwars would start when i mentioned this tonemapping the fact is you can easily mimic fstorm with corona or vray and viceversa but that is just one scenario.
when you have alot of burned areas vivid colors darker shadows ect is where you know the limitations of a tone mapping.

any user of Fstorm can easily create better renderings out of the render without too much experience in this field where we need to use postproduction and 32bit to mimic a better tone mapping.

this is one of the reason why alot of people see better renderings over there plus it has its own glare/flare system well implemented.