colored water ..alcohol, amazingly simple manages to be so frustrating!

[eric_strebel]

Classic example of why I hate the render engine in Fusion 360.

I want colored water, alcohol specifically. I select water change the color to be like whiskey and select alcohol...result...

Grey...no mater how i change the color it stays grey...Something so simple is so hard, hate CAD, something so easy is so messed up hard, did I mention that yet!

uggghhhh, I spend 20 minutes making a bottle, then I spend 1.5 hours trying to get the color of the water...

Still nothing, oh BTW the rendering engine has a bug that when I select custom size it makes my image smaller, how that does not get fixed is beyond me. See images, they are full frame when I start but don't turn out that way.... Seriously frustrated once again!

[PhilProcarioJr]

The air gap method has always been and will always be the worst way to handle fluids in a glass. This has been proven time and time again. I can't believe AD would even suggest this as a viable solution......I don't even know what else to say about it....

[cekuhnen]

@PhilProcarioJr That is why I said that I think that the engineer was simply distracted with more important work but asked to quickly leave a note in the thread here and because he was distracted just gave the opposite and wrong answer.

It happened few times already to me that those rushed answered provided by AD were wrong.

[PhilProcarioJr]

@cekuhnen

Here is why I don't think the engineer made a mistake...

He said "Hi, I am an Autodesk engineer, and I have a few details to add to anavaidya's correct explanation above."

Her method was also to use the air gap method.....and I am sorry to say but her method doesn't look realistic to me at all...but that's just my opinion.

[colin.smith]

I'd like to thank the forum for the interesting discussion.  From what I've been reading, the issue comes down to how you choose to model glass container. I suspect many CAD designers don't think that there is a need to model a scotch glass in a particular way in order to get a good/accurate looking render.  They just model a glass and then afterwards try to make a picture of it with scotch in it. To me that seems counter-intuitive. Also it comes down to the purpose of the image -  to convey design intent, or to be used in for marketing. The requirements for how good/accurate the rendered image looks is not always the same.  

I believe the goal for Fusion rendering should be to provide a way for the average CAD user to get a decent image out of Fusion without having to be a rendering guru - if photo real marketing images are needed that job is better left to rendering professionals using high end animation/rendering software who truly understand how models need to be made, how materials/environments need to be set up and how the rendering engine works. Having said that, I feel we could make it easier to get a better result than we do today.  

Colin

[cekuhnen]

@colin.smith

Few thoughts:

The rendering approach Fusion uses similar to Keyshot is very popular to new users but what new users often dont realize is, is that those two engines with their approach use certain tricks

to achieve certain results. In engines that do not use those shortcuts you do not have to be aware of them and thus can design you shot like you would do in a photo studio.

From a user and student perspective I see very often the approach of this is to complicated and I do not want to learn it. But 3D rendering is 3D rendering and there is and always

will be a certain learning curve to it.

With this I hope you notice that I put the focus of responsibility to the user not the software.

This actually means that only easy is not really true because the new users need to be taught and thus understand how it works.

This rendering with a vessel and fluid is a pretty good example to show multiple factors one needs to keep in mind.

A basic how to render fluid tutorial would be doing the job for the educational component here.

And when then the engine can mesh as perfectly the surfaces so the two common approaches to render fluids in a glass vessel can be ignored that would be incredibly perfect because it cuts the time down.

[colin.smith]

@cekuhnenGood point Claas, there is a balance between software and education for sure. 

I'm sure there are improvements that could be made to the rendering engine to make this better/easier (I'll let the rendering PHDs figure that out) but we also need to be able to direct our users to helpful tips like the ones provided in this thread. 

[colin.smith]

@cekuhnenGood points Claas, there is a balance between software and education for sure. 

I'm sure there are improvements that could be made to the rendering engine to make this better/easier (I'll let the rendering PHDs figure that out) but we also need to be able to direct our users to helpful tips like the ones provided in this thread. 

[PhilProcarioJr]

@colin.smith

Some points I would like to make...

" I suspect many CAD designers don't think that there is a need to model a scotch glass in a particular way in order to get a good/accurate looking render.  They just model a glass and then afterwards try to make a picture of it with scotch in it."

I have found in my career that the customers/employers expect you to be able to do it all. Almost every CAD designer I know wants to be able to do the accurate design, then photo-realistically render their work. Thus the problem you have to understand how lights work, how surfaces react to light, dielectrics, sub surface scattering, reflections, refraction, photons etc.... The human mind knows what looks real and what doesn't so if you settle for a good image it will look CG and ultimately your customer will not be happy. CAD and photo-real are not easy and no matter how hard you guys try their will always be a steep learning curve. You can't make software compensate for lack of experience and knowledge...sorry but it's not gonna happen. One thing that really sparked my curiosity about this thread...there has been no mention of caustics or sub surface scattering...two key elements to fluids and glass. Another case in point, glass will never look realistic without some sort of subtle bump map applied to the surface. Glass is not perfectly smooth so reflections don't have perfect edges. I spent an entire year studying glass, reflections and light. So this might just be my opinion but it was based on real world examples. Now on the topic of fluids, there are so many things to consider viscosity being the biggest factor. The denser the fluid the less light passes through said fluid...enter sub surface scattering. In the areas where your looking through the least amount of fluid it will be more transparent and towards the center which is the thickest part will be more opaque.....anyways enough about that as it can go on all day....  

"To me that seems counter-intuitive. Also it comes down to the purpose of the image -  to convey design intent, or to be used in for marketing. The requirements for how good/accurate the rendered image looks is not always the same."

Forgive me but I need both, and so does the majority of designers I know.

"I believe the goal for Fusion rendering should be to provide a way for the average CAD user to get a decent image out of Fusion without having to be a rendering guru - if photo real marketing images are needed that job is better left to rendering professionals using high end animation/rendering software who truly understand how models need to be made, how materials/environments need to be set up and how the rendering engine works."

Because the Fusion team wants "the average user" to be able to render an image the rest of us power users suffer because you dumb down the UI and hide important controls and access to advanced setups for photoreal rendering. That to me is counter intuitive. See my point above employers and customers want one employee to do it all.......If you are a CAD person that says I will leave the rendering to the next guy then the guy that can do both will steal your job and you will be the one struggling...

On a side note, AD has two of the most brilliant men on the face of the planet in their employ (Jos Stam and Duncan Brinsmead) sit down and talk with the men you might learn a few things. 

Just my 2 cents

[PhilProcarioJr]

@cekuhnen is correct, instead of trying to create the (do my job for me button), make instructional videos and tutorials that show the user how to get realistic results.

[milos.hasan]

I certainly did not say my suggestion (1) is the most realistic, in fact I said it is not physically accurate, but it is the best bet currently unless you want to get fairly technical. The method (2) we are working on will be both physically accurate and easier to use.

Also, your result looks worse than anavaidya's above, so my suspicion is that you did not even do (1) correctly, which just proves it is not yet easy enough. All the more reason to introduce a better method like (2), which is based on this paper:

www.idav.ucdavis.edu/func/return_pdf?pub_id=772

I believe a few commercial renderers already implement this (or closely related) method.

[cekuhnen]

@PhilProcarioJr Judging by the build-in render engine speed you might not want to consider caustics anyway. Even in Cycles which is GPU powered and a path tracer it is slow.

On the other hand Thea is blazing fast with caustics and can use CPU and GPU together. The plug-in is there for Fusion and it works very well with it.

I think the render UI is well solved in Fusion. You have a basic setting and a more advanced setting. My biggest problem concern is simply that it can do even less than Keyshot.

No real lights if you need it and it is only HDRI IBL based which is not really the best approach anway.

The materials Fusion offers I think are pretty great specially also how the 3d Viewport and simulate them.

Because of the painfully slow speed of the engine however I can only advice to use it for basic screenshots renderings during the design phase.

I stand with the point that many who claim rendering is hard simply are too lazy. We learn sketching for years and but then with rendering have no patience...

If you do a little smart phone photography and google some studio set-ups you can easy create pretty good looking renderings and studio set-ups.

It is not really that hard.

[cekuhnen]

@milos.hasan

I did not say that you propose the most realistic solution. I said you propose a completely wrong solution.

First Image: Ok short cut

Expanding the fluid surface is a better shortcut not generating the extra reflections and showing correct refraction among the glass edges.

Second Image: Your proposed solution

Introducing the air bubble is always leading to wrong reflections and refraction.

Do you know see where you are wrong?

Where you are correct is that once the render engine solves the refraction issue with two surfaces being parallel then maybe you can

actually ignore those workarounds and just model the vessel and fluid as they are in reality.

I read through the paper and am quite surprised that they also propose that it is common to add a small air gap. Quite amusing.

[PhilProcarioJr]

@cekuhnen

"I think the render UI is well solved in Fusion. You have a basic setting and a more advanced setting. My biggest problem concern is simply that it can do even less than Keyshot."

Given the fact that you can't create your own PBR materials and shaders I feel that the Fusion renderer is very limited. It makes no sense to me whatsoever to not be able to do this. Keep in mind I don't think you should have to but we should have the ability to if we want. Every time I go to grab an existing material and modify it to my needs it's always missing some input I need. I have spent more time wasted looking for a material that I can modify to fit my needs then I do getting work done. This is why I use substance designer with the IRAY renderer for all my personal work. It's blazing fast and I can create any pbr material I want fast. It gives me complete control.

"No real lights if you need it and it is only HDRI IBL based which is not really the best approach anway."

I agree, the only time HDRI is a great solution is if you own something like HDR Light Studio, then you can create the perfect lighting setup.

"The materials Fusion offers I think are pretty great specially also how the 3d Viewport and simulate them."

I don't feel the same way you do about materials in Fusion, I find them to be sub-par and limiting. The realtime viewport in IRAY is IMHO the best out there and within seconds you have a close to final rendered image...no waiting.

"I stand with the point that many who claim rendering is hard simply are too lazy."

I couldn't agree more with this statement.

[PhilProcarioJr]

@cekuhnen

"I read through the paper and am quite surprised that they also propose that it is common to add a small air gap. Quite amusing."

I found that quite amusing also, and the image they supplied to back up their theory looked very CG and not realistic at all......lol

Oh well difference of opinion I suppose....

[cekuhnen]

@PhilProcarioJr 

Yep that is what I thought as well. It looked and read rather dated. The thing is that the proposed solution is the first I learned years ago lets count I think 20 or so.

I have to say that I do not have access to that particular software.

Realtime PBR shaders are also in my experience never rendering like when using in a finale render engine

thus I am fine with what Fusion creates as they seems sufficient for previews.

The general problem with rendering NURBS surfaces is also for texturing the UV layout. In Maya you can do that quite fine and direct the flow.

In Blender I do that with the UV mapping on a render poly surface. That limits those render engines anyway who cannot give you control over this.

The moment you use a texture material and have something different like a box you hit a wall head on.

[PhilProcarioJr]

@cekuhnen

"I have to say that I do not have access to that particular software."

You really should check it out, its very affordable.

"

Realtime PBR shaders are also in my experience never rendering like when using in a finale render engine

thus I am fine with what Fusion creates as they seems sufficient for previews."

Here are some comparisons between realtime, interactive, and final renders....

"The moment you use a texture material and have something different like a box you hit a wall head on."

I feel your pain on this one......

[michallach81]

C'mon... You treat us like a kids, we're not new in this business. Maybe we're not developers , but it's not the first time we facing that kind of problems. That's why we come immediately with solutions. Sometimes better, sometimes worse, but still solutions.

We support our arguments with images as a reference. We even add files, so anyone could easily argue opposite.

Because of some weird corporate etiquette your finding "sophisticated" arguments like "unless you want to get fairly technical", to defend failed arguments. Your first method, same as second require 3 steps, 2 revolutions, and 1 offset. My is equally "technical" 2 revolutions and 1 offset... All tree require to know about them in a first place, which make all of them not "out of box".

If you think we're wrong, prove us. Show image and attach file to review.

[cekuhnen]

@PhilProcarioJr Pretty cool 300 $ per year for GPU rendering

Alias Portfolio and similar apps run at 10k a license 😉

[anavaidya]

While we implement a rendering technique that can handle this situation, we also want to keep this in mind - The model (of the glass container) should be physically accurate, so that it can be used for manufacturing. We don't want a user to create a separate model just for rendering.

Here is a test we did using 3 methods of modeling and rendering a liquid in a glass container, side-by-side. 

1 (left). Modeled the glass as it would appear in the physical world. Modeled the liquid inside it, with a slight air-gap. As discussed earlier, this is not ideal but doesn't break the model and gives a reasonable result.
2 (center). Model prepared for nested dielectrics. Here the liquid slightly intersects the glass model. Currently it does not render correctly but it will once we implement it, it will work. This is the method described in the paper attached earlier. The approach also requires the setting of priority parameters. But we'll get to it once we are closer to release. Something to note is that this method is also used by Renderman, Maxwell and Arnold. This is the method that we are going for.
3 (right). Modeled such that the glass object is split into 2 parts, along where the liquid starts and delete some surfaces (as described earlier in the thread by cekuhnen). It gives a more accurate render, in this case. But you need one model for manufacturing and separate one for rendering, which is not what we want. Also, it may not work for certain complex surfaces.

[cekuhnen]

@anavaidya

This is getting a funny note now, do you realize that?

Instead of shrinking the fluid surface just expand it.

This is not really that hard you know 😉