Poll

What option do you prefer to control hair highlights size?

Glossiness (currently used, also behaves similarly to CoronaMtl)
Roughness (usually hair shaders/materials use roughness instead of glossiness)

Author Topic: 1.7 - Hair shader playground and feedback  (Read 41981 times)

2017-07-12, 15:56:26
Reply #60

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com

2017-07-12, 21:05:29
Reply #61

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
Bigguns, we cannot insert 0 as Curve subdivs. with 1, I have this strange white dots on every vertex.
I think it's a bug. also, you wrote to put "zero" to avoid the white dots. But.. in which option =) ?

And, I have discover that "Curves subdivs" works also with H&F =)

I meant, put zero in the glint, in the shader , by default the glint value is 0,2 , put 0 and it will be ok .

2017-07-12, 21:44:46
Reply #62

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
some more old test ^^

2017-07-12, 22:27:14
Reply #63

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com

2017-07-16, 10:51:29
Reply #64

FrostKiwi

  • Active Users
  • **
  • Posts: 686
    • View Profile
    • YouTube

Did an animation of a furry teapot using H&F and found some corona buginess:
  • CoronaDr does not transfer .stat files of the H&F Precalc and instead precedes to merge frames as of they had no precalc with ruined images as a result. Network mapping the Statfiles worked fine
  • Both Corona normal, Corona DR and backburner never unlock the Ram used by Corona. It keeps growing and growing with each Frame until it runs out of RAM, then it pages. Paging is handled fine with no performance loss, until it is maxed out. Then corona crashes and backburner saves a black frame, then the whole process is restarted.
  • normally my 8350@4,7ghy does ~2.2mio rays/s in normal scenes. With CoronaHairMTL in mrShader slot it crashes to 600k rays/s. The interesting thing though: if you convert to mesh, it does 900k rays/s. No opacity used.
    Should there be really such a performance hit for not converting to geo in 3dsMax before sending it of to Corona?
Animation as MP4 in attachment

Edit: screenshot in Attatchments of Corona massing Ram and never unlocking until crash.
Both DrSlave and Workstation produced a black frame once they ran out of all virtual memory.
Screenshot is after crash an hour before. Where it was 45 /45 GB virtual ram.
This is after ~8 frames. Each frame adds ~1.5gb ram that never unlock
« Last Edit: 2017-07-16, 11:06:11 by SairesArt »
I'm 🐥 not 🥝, pls don't eat me ( ;  ;   )

2017-07-16, 12:16:58
Reply #65

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
personnaly I did not tried it with hair and fur since I don't use it anymore.. it's so so bad.. but , you should not apply the material to the material slot in the hair and fur, it's for MR material. Just appy it to the emitter mesh , that's it. Also, in the environment ( never render the hairs as geo ! that shader is not made to work well on geo hairs! you'r loosing most of the quality of it by oding that and it takes more RAM ) I guess you should use buffer mode , but I'm not sure since I did not test it with hair and fur.

2017-07-16, 12:24:15
Reply #66

FrostKiwi

  • Active Users
  • **
  • Posts: 686
    • View Profile
    • YouTube
personnaly I did not tried it with hair and fur since I don't use it anymore.. it's so so bad.. but , you should not apply the material to the material slot in the hair and fur, it's for MR material. Just appy it to the emitter mesh , that's it. Also, in the environment ( never render the hairs as geo ! that shader is not made to work well on geo hairs! you'r loosing most of the quality of it by oding that and it takes more RAM ) I guess you should use buffer mode , but I'm not sure since I did not test it with hair and fur.
Ohh snap! Thanks for the tipp. Gonna retry
I'm 🐥 not 🥝, pls don't eat me ( ;  ;   )

2017-07-16, 22:59:15
Reply #67

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
corona actually uses the mr slot instead of emitter material
Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2017-07-17, 13:44:40
Reply #68

FrostKiwi

  • Active Users
  • **
  • Posts: 686
    • View Profile
    • YouTube
corona actually uses the mr slot instead of emitter material
Yeah, I was about to say after I double checked with the Changelog.
I indeed had it set to geometry and that caused all the incorrect Ramlocking. With buffer mode it seems to.be all fine.

However, see picture attached. 2.5mio hair causes 500k rays/s, but transferring to mesh gives a boost to 800k rays/s (in normal scenes it is usually 2-2.5mio)
Is this subject to change, or will H&F buffer Hair always be slower than convert to mesh ? (Ignoring the horrible Ram explosion after mesh convert)
Also will parse times always be this aweful with H&F? For 2.5mio strands its almost a minute.
I'm 🐥 not 🥝, pls don't eat me ( ;  ;   )

2017-07-17, 18:06:30
Reply #69

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
corona actually uses the mr slot instead of emitter material
Yeah, I was about to say after I double checked with the Changelog.
I indeed had it set to geometry and that caused all the incorrect Ramlocking. With buffer mode it seems to.be all fine.

However, see picture attached. 2.5mio hair causes 500k rays/s, but transferring to mesh gives a boost to 800k rays/s (in normal scenes it is usually 2-2.5mio)
Is this subject to change, or will H&F buffer Hair always be slower than convert to mesh ? (Ignoring the horrible Ram explosion after mesh convert)
Also will parse times always be this aweful with H&F? For 2.5mio strands its almost a minute.
do you use "render as planes" option? unchecking it slow things down pretty considerably. Generally usually hair is a bit slower to intersect, but way less memory hungry
Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2017-07-17, 19:52:05
Reply #70

FrostKiwi

  • Active Users
  • **
  • Posts: 686
    • View Profile
    • YouTube
Quote from: Ondra link=topic=16730.msg105691#msg105691
do you use "render as planes" option?
Yeah, it is on by default. Was just suprised to find mesh faster than Hair & Fur Buffer.
I'm 🐥 not 🥝, pls don't eat me ( ;  ;   )

2017-07-18, 07:58:14
Reply #71

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
corona actually uses the mr slot instead of emitter material
Yeah, I was about to say after I double checked with the Changelog.
I indeed had it set to geometry and that caused all the incorrect Ramlocking. With buffer mode it seems to.be all fine.

However, see picture attached. 2.5mio hair causes 500k rays/s, but transferring to mesh gives a boost to 800k rays/s (in normal scenes it is usually 2-2.5mio)
Is this subject to change, or will H&F buffer Hair always be slower than convert to mesh ? (Ignoring the horrible Ram explosion after mesh convert)
Also will parse times always be this aweful with H&F? For 2.5mio strands its almost a minute.


Nice teapot haired ^^

2,5 millions of hairs.. less then a 1 min to parse.. not that slow to me. it's 2,5 millions ! give it a chance lol

2017-07-18, 09:19:32
Reply #72

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
An other test


2017-07-18, 10:26:22
Reply #73

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Hey guys,

I just want to react about the poll of this topic. No matter the outcome of the poll, I think that what matter the most is that all the shaders follows the same rules. Thus, if the coronamtl stick with glossiness input, then the Hair shader should behave the same way. Btw, from these two workflows, one of them seems to stand out as the standard today and it's the roughness one. I think YOU (devs) have to make that choice, not us (users), because i think (in most cases) you are responsible for making adopt these new workflows by imposing them. I think that's how a standard become a standard. I think it's really great to ask user feedback and this is one of the thing i love with corona development but on certain subjects like this one, I think we should not intervene.

Of course, this my point of view and i will understand some people will disagree with me. I just want to introduce some constructive discussion about that because it seems important to my eyes.


2017-08-08, 20:05:38
Reply #74

Bigguns

  • Active Users
  • **
  • Posts: 114
    • View Profile
Hi guy's , my latest hair tweak, added custom glint effects and hair breakdown into the reflection . Look's quite real now :P Lit with an hdri

« Last Edit: 2017-08-08, 20:40:26 by Bigguns »