Fstorm render

[joconnell]

Cool!

So photoreal on the vray side means implementing two things perhaps - better glow / glare / bloom and then some photographic tone curves. The hard part of it is going to be getting the artists making the scenes to make photorealistic scenes in terms of their lighting, materials and camera settings - tone curves and glow on a render that's using unrealistic ingredients will only give you a slightly different looking unrealistic render, it's not going to be the part of the process that makes the biggest difference.

The twitch stream thing is a great idea though!

[wyszolmirski]

What's wrong with Glow in V-Ray ? I'm not using it much to be honest but didn't find it lacking anything. Just curious what's the story with it.

Two things I would like to have in V-Ray:

1. Additional tone mapping as I find Reinhard to be good but I would like to have other options as well.
For some time I use filmic tonemapping in Nuke and I'm happy with results:
http://www.dabarti.com/screens/s3d8c...c8274b396t.png vs Rainhard: http://www.dabarti.com/screens/s5a05...704b809203.png and vs linear http://www.dabarti.com/screens/s8598...b3717cf42G.png

While having that in VFB, I would probably don't color grade so much in Nuke / AE. We don't add motion blur or DOF anymore in post due as it's faster and better to render it in RT.

2. As you still have to be able to correctly compose the shot I would prefer if V-Ray wouldn't bake the color grading into RGB_color while saving to separate elements, instead save it as separate file like RGB_color_VFB or do it other way around (to not break current workflows) - bake grading to RGB_color but save pass RGB_color_ungraded as well.

Best,
Tomasz Wyszolmirski

[jstrob]

I have yet to really experience the benefits of RT. I upgraded my video card, thinking I could start to use it, but that just hasn't happened. When I tested FStorm, it just worked, so I get that. Yes, the images are good, but not any better than anything else, or are you saying you think that they are?

yes they are. I'm now sure of it. And i have a better idea why also. Better glare, look up tables and better default settings (GGX, maybe something in the glass). There is something in the shadows, the reflection the glass that is a bit more photoreal by default at least. And it is probably all linked to the look up table and glare being more photoreal. Even in post with expensive filters I can't think of any glow or glare effect that can easily look as photoreal as the fstorm one.

[jstrob]

What's wrong with Glow in V-Ray ? I'm not using it much to be honest but didn't find it lacking anything. Just curious what's the story with it.

Two things I would like to have in V-Ray:

1. Additional tone mapping as I find Reinhard to be good but I would like to have other options as well.
For some time I use filmic tonemapping in Nuke and I'm happy with results:
http://www.dabarti.com/screens/s3d8c...c8274b396t.png vs Rainhard: http://www.dabarti.com/screens/s5a05...704b809203.png and vs linear http://www.dabarti.com/screens/s8598...b3717cf42G.png

While having that in VFB, I would probably don't color grade so much in Nuke / AE. We don't add motion blur or DOF anymore in post due as it's faster and better to render it in RT.

2. As I you still have to be able to correctly compose the shot I would prefer if V-Ray wouldn't bake the color grading into RGB_color while saving to separate elements, instead save it as separate file like RGB_color_VFB or do it other way around (to not break current workflows) - bake grading to RGB_color but save pass RGB_color_ungraded as well.

Best,
Tomasz Wyszolmirski

I think your 3 images shows exactly why the fstrom facebook group has so many photoreal renders. The look up table. I know it is something that should be more in comp. but if we can see it directly in V-Ray frame buffer that's even better. In fact maybe we can I just didn't know really much about LUT before discovering fstorm.

Also very good idea to have more options to output with and without the baked color grading. And it was brought before but also some elements like multimatte and zdepth should never be color graded anyway.

[Moriah]

But only 2-3 days ago fstorm included LUT support...

[vlado]

So photoreal on the vray side means implementing two things perhaps - better glow / glare / bloom and then some photographic tone curves.

You've no idea how very depressing that sounds All that work to make sure GI, lighting, materials, reflections etc are all accurate and fast don't matter, but apparently what matters is a bit of lens effects and tone mapping. That's life, I guess

The hard part of it is going to be getting the artists making the scenes to make photorealistic scenes in terms of their lighting, materials and camera settings - tone curves and glow on a render that's using unrealistic ingredients will only give you a slightly different looking unrealistic render, it's not going to be the part of the process that makes the biggest difference.

That's what I've been trying to tell you guys - no amount of glare or color grading will turn a badly set up scene into a masterpiece... V-Ray is mostly set up for photoreal rendering by default, but it still amazes me how people manage to purposefully get flat and uninteresting renders through wrongly placed bounce cards, additional hidden ambient lights etc.

The twitch stream thing is a great idea though!

It's an interesting idea, but how would that be different from the forum (other than being live, obviously)? Here at least we have time to think through the replies and provide some kind of factual proof for various statements, which is not possible in a live session.

Best regards,
Vlado

[jstrob]

You've no idea how very depressing that sounds All that work to make sure GI, lighting, materials, reflections etc are all accurate and fast don't matter, but apparently what matters is a bit of lens effects and tone mapping. That's life, I guess
Vlado

Don't forget that we have only seen simple archviz or product scenes so far from fstorm (no displacement, hair, volume etc). And the reason is that it just can't do anything else for now and will probably take years before it can do something else. It is not just a detail that matter most, it takes all the details put together and then you get photorealism. I'm sure chaosgroup can add a small detail or 2 that fstorm has and stay the king of renderer!

Why GGX is not the default BRDF by the way?

[LudvikKoutny]

I really wouldn't say V-Ray in its current state is not photoreal. I'd actually say the opposite.

The difference lies in just a small bunch of different default values, and a bunch of features that have not been refined yet. But the thing is that while these differences won't make images of inexperienced people (or people with bad taste) look better, it will help those, that are experienced good artists, just do not invest much time into digging into technical stuff and browsing forums to find hidden tips and tricks.

Honestly, I think that when it comes to achieving ultimate photorealism... every single effect that puts together a real photograph is equally as important. Historical progress of rendering development has led us to a false impression that there is order of importance when it comes to how much shading and lighting effects impact resulting photorealism of an image.

For example we think that true area lights instead of sharp point lights with shadows are more important than having GI, that having diffuse GI is more important than having caustics, that proper raytraced glossy reflections are more important than correct fresnel reflection falloff, etc...

So then we tend to think, that for example disabling casting shadows on a random small object in our scene will be a disaster, where as having a missing glare star on a directly visible sun disc is just a minor thing. Or that if we do not make some material to be correctly reflective, it will draw attention a lot more than a missing caustic.
Or when you have only 2 bounce Global illumination, it often doesn't matter anyway, because many people usually abuse the hell out of a contrast button in post process phase anyway, so they will make even 25bounce GI like 1 bounce after the adjustment, where as if you miss a glow around overexposed window, your eye will pick up on that more likely.

Well, I think that human eye will pick up all these missing things equally, that sharp caustics are as important part of the light light transport as basic shadows, and that glares and glints are equally as important as materials having reflections and so on. Only when you have all of the puzzle pieces at your disposal can you put together entire picture, and more of the pieces you have, the closer to the entire picture you are.

I bet you that if you take any scene, and do a version A, where you will do really poor rushed job on shading and lighting, but then do a great job on post processing and secondary optical effects (glow, glare, vignette, aberration) and then do version B, where you spend a bit more time getting shading and lighting right, but won't do any post processing work at all, if you look at both images side by side without spending too much time staring at them, you won't really think one looks significantly closer to photo than the other one. They will both have something about them, but none of them will convince you even for a second it's a photograph

[joconnell]

You've no idea how very depressing that sounds All that work to make sure GI, lighting, materials, reflections etc are all accurate and fast don't matter, but apparently what matters is a bit of lens effects and tone mapping. That's life, I guess

Best regards,
Vlado

I don't care about them personally, just trying to interpret what some people are asking for or what they think "photographic" is and how that'd translate into features that could be implemented to make the users that want a finished look out of the frame buffer - If it's any comfort sometimes I'm horrified at what happens to my carefully balanced lighting when it goes off to a compositor and all of the relationships between light, shadow, reflection and spec suddenly become open to interpretation and personal taste!

On my side I'm all about accuracy - at the minute I'm trying to reverse the tone mapping curve of my stills camera so I can get linear results so I'm trying to shoot macbeth charts, munsell neutral value charts and then profile the raw files I get from the camera to remove any of the photo look / tone mapped stuff so I'm getting accurate brightness and hue values from it - In fact I'd love to talk to whoever did the hardware calibration of your scanner to see how they profiled whatever camera or sensors are inside of it to remove any colour profiles or casts from the lights

I get you on the factual stuff too, language can be very difficult when choosing the precise wording for something to put a correct idea across and leave no room for any other interpretation

[Nicinus]

The difference lies in just a small bunch of different default values, and a bunch of features that have not been refined yet. But the thing is that while these differences won't make images of inexperienced people (or people with bad taste) look better, it will help those, that are experienced good artists, just do not invest much time into digging into technical stuff and browsing forums to find hidden tips and tricks.

Couldn't agree more.

[PIXELBOX_SRO]

Any idea what Vlado means by the wrongly placed bounce card? Sorry for my ignorace, maybe its just the translation issue but i have no idea what he is talking about.....portals?
And please dont laugh if i am missing somethign very obvious ��

[PIXELBOX_SRO]

Nevermind i know what he means now ��

[mitviz]

i think the lighting and camera settings are very minimal worries though, biggest thing that seem to bring realism is the models themselves and material because in real life there is sometimes terrible lighting and when you use a camera you can change the lighting as you like with a professional camera but the object and things around will stil look realistic now matter what so its something i have wondered for awhile and tested it with my camera, so more work placed on the part of the material and models makes the biggest difference and then that little extra help from the renderer in the end takes it to the next level of realism. again that is subject to me and my observations. by the way am using the cannon D1000 with a few types of lenses

[PIXELBOX_SRO]

for anyone who wants to play with tonemapping and LUTs you can try the VFB+ plugin it even does DOF and lens effects.
That sould give anyone plethora of options to achieve the look they want.


I dont think things like lens effects etc shoudl be key features to focus on if there is just so many other plugins outthere that can také care of this nicely. I am glad the Chaos guys are focusing on the IPR and RTGPU thats much much more important than some grading tools.

This reminds me of a friend i have who bought BMW instead of AUDI because he like one particular tiny detail in the interior although AUDI has offered him a lot more than BMW.
So i feel Vlados pain when topics such as these are brought up.

FStorm is a nice renderer for sure, the resilts are great but its in its birth age and before its somewhere where its usable for a production studio, Vray will be again miles away from it.

[Nicinus]

for anyone who wants to play with tonemapping and LUTs you can try the VFB+ plugin it even does DOF and lens effects.
That sould give anyone plethora of options to achieve the look they want.

Unfortunately VFB+ is apparently not compatible with the denoiser.