Biased Vs Unbiased Rendering
Biased: Gives the user complete control over how the renderer interprets the scene. This means calculations for Spectacularity, Global illumination, Final gather, Shadows etc. have to be manually entered by the user and are a result of how the user feels the lighting interacts with the objects within the scene. This doesn’t always mean that the users interpretation of how the light interacts is physically correct, this can take a lot of finessing and if not done in the correct order, can quickly look “wrong”. Renderers such as Mental Ray, Vray and Renderman are known as biased renderers.
Unbiased: Is the opposite of a Biased, meaning that the renderer is in control of calculations. By doing it this way the renderer can create physically correct renders with minimal work for the user, who is in control of noise clean up and the amount of bounces in Diffuse and Refraction. This offers more control and less room for error with calculations. Renderers such as Arnold are unbiased and are increasing in use within the film industry due to their speed and render quality.
The pro’s and con’s of each render type can be broken down;
Pro’s; More control, faster render speeds, ability to choose quality of work also the user can predict possible render times, render times can go from seconds to minutes depending on global illumination and final gather photons as well as light quality and the amount of objects within the scene, better for use with hard caustics.
Con’s; Less ability to achieve realism and this isn’t dependent on how much quality the render has, but how the render handles information. However it can still produce quality renders.
Pro’s; Physically correct, realistic, less work for the user meaning some adjustments regarding camera and scene clean up and real time updates using the IPR render option, better at soft caustics.
Con’s; The render is never actually done, render times can be unpredictable and the overall quality cannot be chosen.
For the final part of this assessment, we were given “client feedback” to make our adjustments to the tractor beam simulation before final submission to try and replicate what might happen in industry. Below is the feedback I received for my previous blog post;
The outer cone of the tractor beam has a nice look to it. I like how it has the diagonal shafts coming from the top, it gives some variance to its makeup.
The tractor beam is turning on too slowly though. It needs to happen much faster so that there is time to lift the mannequin into the UFO. Think no more than 15 frames.
We need to get some more detail into the simulation. I’d like to see thin rings run along the surface of the outer cone. These would move from the UFO down to the ground. As they run along the outer cone they would leave a faint trail of particles that would waft away from the outer cone and dissipate into the air. When the rings hit the ground they would dissipate and fade off.
I also want to see more detail in the current outer cone that you have. This will probably mean you’ll need to make the particles smaller and have more of them.
The colour is close, but it has too much green in it. Try to make it more of an electric blue.
Lose the grass type things that come up from the ground. Also, make sure the tractor beam is timed to the opening and closing or the UFO hatch doors. At present that is not happening.”
So I approached this by increasing the over samples of the particles to 3, to make the main part of the cone seem more like a veil than like individual particles. By doing so as well as trying to increase the time of the tractor beam contacting the earth, I lost the cross over effect of the particles. I do rather like the electric blue colour more than the aqua colour I had previously.
I had a bit of trouble trying to get the little rings to look how I wanted them to, I tried several different fields, emitter types and also goaling to an object. In the end I emitted particles from the initial rings I created and turned the visibility off, because I found if I had used any form of keys in the visibility attribute I had issues with the particles turning on and off again. So these initial rings became my emitters for the visible rings. After trying several fields, Newton, Gravity, Turbulence, Air, and Drag, I settled on using a Radial field. This still didn’t quite give me what I was after but I needed to keep moving forward in order to finish the simulation. I would rather that the particles actually moved away from the centre of the rings in a horizontal motion and slowly glittering into nothing.
I removed the particles floating up from the ground, I can see how they weren’t the desired effect. I only kept the point particles falling from the UFO, these I changed to a golden colour to add a little subtle contrast to the simulation and added a noise texture to the opacity ramp to make the particles sparkle on their way down.
After struggling a lot with the rag doll simulation, I finally got the simulation to work however I had further issues importing my file into the VFX file, so to get around this problem I referenced the original reference and made the hinge constraints and parenting constraints within the VFX file. I know this might not be the way it was meant to work, due to the assessment reflecting industry, however I wasn’t sure of the issues and needed a quick solution, so I improvised.
Below is a playblast of my final simulation, I feel the ring particles could be a little too fast but this can be changed later if need be.
This post is the next lot of self critique and a breakdown of processes I went through to achieve the current version of my tractor beam.
Between my concept post and now, I have found one more reference I would like to add to my research. Recently the cinematic trailer to Overwatch was released, there is a part within the trailer where there is a close up of a glove in a display case. The particle effects used here are similar to the idea I was going for when I concepted my core beam.
Currently the core beam is using both Particles and nParticles, I made this decision once I played around with a few options using Particles for the main cone shape. My first solution was to create a cylinder, adjust the verticies to suit the shape of the tractor beam and emit from object. I had issues here where I couldn’t get the particles to do what I wanted to, the opacity was either too much or not enough making it really hard to see through the “veil” to see the wooden man. Also finding that it was hard to find “the right amount of particles”. Below is a screenshot as an example of the result using this method.
My second attempt at a solution, was to create a volume emitter in the shape of a cone. This I had some trouble with, trying to get the emitter to release the particles how I wanted them to. I struggled with trying to get the cone shape wide enough, I am sure there was something I wasn’t doing and it would have been an easy fix but in the end I had a feeling I would have ended up with a similar result to the first method.
The final solution I tried was to use nParticles instead of the old particle system, I found this was a quicker, simpler solution to the other two methods. Using two torus’s, I emitted from the top one and made the bottom torus a goal weight. I found it easier to control the opacity and have the particles look more like a “veil”, one thing I would do is to get the core beam to spin in a slow clockwise motion as I feel the accidental cross over motion at this point in time feels too static.
The other particle systems that are within the scene, are made from Maya’s original particle system. Using a similar technique to the nParticle solution, I used torus’s to emit from. The only thing I feel needs to be tweaked is the particles being emitted from the ground. I feel the particles at the moment are too bright, they could be toned down just a little bit by changing the lightest colour in the ramp on the Opacity PP to a darker shade. I am not quite sure if I should add an additional light source underneath the wooden man…
At this point in time I am thinking to scrap one of the elements of the core beam, this I can add in at another time when I can play around with the settings and explore the best way to achieve the desired result. This part is the 5 particles that drop down from the UFO as a goal to the wooden man to appear as though these particles are lifting the wooden man back into the UFO.
So for this post, as the title states, this is a self critique of my progress so far.
As for the progress side of things, my self critique would be that I would like to be further than where I am at this point in time. However that is not the point of this blog so I will work with what I have currently to fulfill the requirements.
I tend to get carried away with the research and development side of things as this is a part of the pipeline I enjoy the most. A lot of my time can be seen as “wasted” here because it’s like a domino effect in my head, so for future I need to really focus on setting a time frame for this task and sticking to it, otherwise it becomes a runaway train. In saying that though, the amount of research is a little overkill but I feel it helps to really understand what it is I would like to achieve within a project.
Concepting stage, I would have liked to get a few more idea variations on paper before hitting Maya. In a way though this is a gift and a curse, a gift because this means time was saved and not too much over thinking was done, diluting the initial idea. A curse however because the first idea might not always be the best idea in conveying the concept asked for by the “client”. The idea I do have in my storyboard, is pretty much where I want to go with my particle simulation.
With what I have done so far in terms of particle simulation, I have only really focused on the “curious” particles, which isn’t very much. At this stage, the speed is far too slow however I will adjust this once I get them working how I want them to. I am aiming to use the second locator as a goal object for the particles to be attracted to, to give them the “insect swarm” sort of characteristic making them seem more organic than the other particles in the scene. The type of particle I use is also not so clear to me yet, currently they are set to points but I will need them to be a bit larger to be effective.
For the tractor beam itself, I am thinking to use a nurbs circle (visibility turned off) as an emitter, one at the entrance to the UFO and another on the ground to create the “meeting in the middle” sort of feel. The characteristics of these particles, I am aiming for them to be like falling snow, soft and gentle. Fields could be added along the way to help give it a bit of variation, however it will all depend on the look once the particles are in place and working within the scene. The particle render type I am thinking to use for this is streaks, I know streaks don’t really have the effect of snow, but I would like to put a tiny hint of a tail on the particles which will have a ramp within the pp opacity node.
The cone shape I was thinking to use a spotlight with light fog. I feel there could be a better way to do this as I would like to have a slight cloud effect within this part of the tractor beam. I did find when I tried to use the spot light (currently turned off in the play blast) that it was harder to control in the way I wished to use it. So my next step will be to try a particle simulation as a solution for this problem.
As for my current play blast, I will try to keep doing some little play blasts as I go to show progress and keep up with the self critique.
The rendering assessment, required us to do two separate lighting scenarios. Here are two images (both are not mine) oh lighting scenarios I liked the feel of, the first is an early morning or late afternoon sort of feel, which if I had time I would have liked to play around with light fog and bokeh lighting effects. This scene has a more natural feeling to the light, it’s quite warm and inviting.
The second lighting scenario is a cooler more artificial light kind of feel. I also like the style of the kitchen in this image and that the cool light is broken up briefly with a warmer light along the top of the bench.
Here are my initial lighting setups; The first one I decided to use the physical sun and sky node within the render settings, I felt this might give me a more natural “early morning” feeling and the sky reflection on the tiles if incase it was to be seen in the final render. I also used an area light to the left of the scene to fake the light on the fruit bowl a bit better as the physical sun and sky wasn’t doing the best job at lighting up the room. There is also another area light to the right of the camera to help soften the shadows.
Here is the second lighting scenarios, initial setup. I only used two lights in this scene, an area light up near the ceiling to replicate the blue, cold, artificial light and a second area light with a sphere light shape on it directly above to imitate the warmer artificial light.
This time round with my pipeline I decided to wait with turning on the photons option until actual textures were on the objects. All of the fruit, minus the grapes were textured in Mudbox, the grapes however were procedurally textured inside Maya using a custom Python script from one of the teachers to create a shader network which flowed through a ramp to randomize the colour of the grapes by only using a single node to control this.
However this shader network connects to a lambert which doesn’t hold the attributes needed to create sub surface scattering (SSS) or control the glossiness. To correct this I needed to plug in the following shaders into the randomized network. The ramp plugs into a misss_fast_shader which controls the SSS, while a mia_material_x shader plugs into the diffuse of the misss. This now gives me more control of the actual fruits texture, I also plugged a mountain texture into the colour on the mia shader to replicate the little brown spots on the grapes. I had to make sure the snow attribute was turned down to a mid grey to keep the green colours from the ramp from washing out.
I decided with all the fruit textures I wasn’t going to tough the roughness attribute as I found this made the “natural” shadows originally cast by the light dull and lifeless. For the other fruit I only really concerned myself with the reflection, glossiness and gloss sample nodes, as well as plugging in the bump and displacement maps.
For the wall tiles I used a marble texture I made earlier in the year and created a bump map using noise to give the tiles a bit more of a texture. The bench top is a procedural texture, consisting of a fractal for the colour and pattern with a noise in the bump map. The fruit bowl itself is a free texture from CGtextures.
Now that I have the textures in place, I can start to readjust the lights and finally add in the photons, final gather, and global illumination. I found working this way was a lot faster and more efficient than the last rendering assessment. I remembered to use light linking this time, and it was crucial I used it on the grapes in the first scene as the python script shader doesn’t respond well to the physical light and sky node, a third light had to be made just to light the grapes, this was another area light with quadratic decay and photons.
The rest of the scene was mainly lit with the physical sun and sky, where I adjusted some of the settings like I upped the Haze to 0.500 so the light had a more diffused “morning” feel to it, turned on the photons as well as finding how to change the quality of the shadows took some time to find as I originally thought I might find these in the render settings, however it is hiding on the second node created by the physical sun and sky button. Here I ramped the shadow samples up to 64 and changed the softness of the shadows from 1.000 to 10.000 this then gave me the final result you see below.
I also adding in some depth of field to the camera, with an F stop of 2 and the focus region scale also at 2.000, this softened the background just enough to keep the viewers attention more on the fruit rather than the background.
For the second lighting scene adjustments, I imported the lights from the second file into the first so I could save some time with the settings of the textures. The only real adjustments I had to make here were to make sure the light above, the intensity was high enough to light the scene but low enough to not blow out the SSS on the grapes as I found when I went too high the top few grapes would turn a blueish grey. The other thing I changed was the intensity of the blue light and used it more as a atmospheric fill light, leaving the bulk of the light to come from the area light above, which also has the photons turned on.
Final touches included the rotation of the camera view and adjustment of the depth of field attributes (only the distance as the F stop and focal region scale remained the same).