I will treat this as a new and separate blog for the UFO dynamics process diary ‘after week 9 feedback’ so please see the week 9 entry for previous blog stuff.
Firstly, here is a summary of the teacher feedback from the week 9 blog itself –
Feedback – liked the movie documenting of experiments.
Action – cool, I will continue with these.
Feedback – good amount of reference gathered , more moving reference would be useful.
Action – Ok, will continue using this reference and address more specifically the movement of the particles.
Feedback – more detail in regard to problem solving and workflows/methods.
Action – sweet, I love talking about this stuff! Putting thought into the ‘how and why’ is one of my strong suits!
Now to the dynamics work.
First of all, here is a recap of what I submitted in week 9 –
As you will notice , I had separated the tractor beam into 3 segments to improve the communication of feedback.
I will talk about theses 3 segments, how they were initially created and the changes I have now made based on –
– the ‘client’ feedback
– timing and distance changes made to brief (previz)
– other changes and embellishments based on my own improving skill-set.
Core Beam –
At the week 9 review, I had created an outer cylindrical beam pulsing up and down rhythmically. I called it the ‘core beam’ as it was the main shape for the tractor.
The positive feedback was that it ‘turns on in a pleasing way’ and the color is ‘bang on’.
There was 2 points of negative feedback –
1. The movement up and down looks weird and is not required.
2. It needs some variance in its makeup, it feels too consistent.
The essence of how the core was built was a poly cylinder –
I needed to modify the geo so that the tractor matched the new scene (the ship is now higher up in Y)
I had used a ramp mapped to the texture rate – ’emit from dark’.
Notice above that the ‘selected position’ is keyed – I needed to change the animation so that the particles just traveled downwards i.e no more ‘pulsing’ and then stop emitting just before the door closes –
which gave me this look of wispy splines creeping down the column – I find that this method of assessing the look of the particles at a texture level, before you generate particles is really helpful – you can see the changes on the fly and scrub the timeline to see how it animates. Then you have an idea of what will happen when you emit – what kind of particle count you will need and how long they should live. With this kind of look, I knew that the lifespan should be very short, otherwise it would just be a mess.
I also decided to render this breakup separately from the main column so I can control the look in comp.
These 2 elements form the new and improved ‘Core Beam’ and it looks like this so far-
At the week 9 review, I had created a circular shaped disk that travels up and down the core beam to add another level of interest.
The positive feedback was that it looks ‘cool’. It ‘adds some movement to the beam which works.’
There was 2 points of ‘negative’ feedback (not that negative, just changes to work on) –
1. ‘It would work better if the fractal just went in one direction. So make it move from the top to the bottom and then die.’
2. A couple of creative suggestions –
a. It might be worth trying this, when the fractal particles hit the ground they turn into mist or some other cool effect. I suggest making it very subtle.
b. Also, it might be nice for the fractal particles to generate a fine layer of particles that sheer off from the core beam as it passes over the top. This could be used to create that dissipating/vapour look that appears in a lot of the reference.
This ‘fractal circle’ of particles will be comped over the existing core beam. It is created in a similar way, using a shading network that is animated to match the timing of the door opening/closing.
So firstly, I corrected the timing to match the new timing of the door. Then I changed the ramp keyframes so that the circles are never visible traveling in the upward direction. Instead they appear to die –
With regard to the creative suggestion of making them turn to mist as they disappear– I thought of a way to do this –
I decided to make the fractal circle particles emit their own secondary particles. I could control the emission rate to just emit at the death of the fractal circle – then these secondary particles sheer away driven by gravity and turbulence fields – they peel away from the main column and fall. I would need to give them just enough lifespan and opacity PP to fade away as they fall.
This required lots of tweaking to get the right look, but in the end I think it works quite well.
In total, these ‘secondary fractal circles’ are formed 3 times at the base of the column and once more at the top of the column just before the door closes. If I have interpreted the feedback correctly then I think this is exactly what the ‘client’ is after –
I was pretty disappointed with what I had achieved with the tendrils in the week 9 submission. I knew what I wanted and what the client wanted but in all honesty I was a bit lost for how to make it happen.
The feedback was that “the tendril feels a too wavy. It feels like the movement is coming from the top, up in the hatch. The head of the tendril, which is the part that moves towards the mannequin, should be the part that leads and determines the movement. What lags behind is more like a trail. Basically the opposite of what you have now. I would like to see 2-3 tendrils that latch onto the mannequin”
This feedback actually made me realize why the movement was wrong – this is how it was created-
Above – original week 9 setup- I had created a spiral shaped nurbs curve. The emitter sat at the top, a sphere was connected to the curve via a motion path and the particles were goaled to the sphere.
You can actually see from this screen shot above what the problem was – the ‘fuzzy tail’ was leading the action which is wrong- the tail should more like a trail.
I tried swapping the action, i.e with the emmiter traveling down the curve. I parented the emitter to the sphere. I also tried adding more sweeps to the curve and deforming the curve with a lattice deformer to get some ‘tornado’ feel. While the theory of this was better, it didn’t really work well enough-
Above – failed solution to tendril issue.
I decided that it would be more manageable if I went with more of a geometry driven solution. This is the method I came up with. I was worried that I was over-complicating things but it actually didn’t take that long to set up-
The basis of this new set-up is
– a sculpted cylinder geo coming to a point at the tip.
– a joint system.
– spline IK added.
– clusters for manual animation.
– fixed position sine deformer to with animated wavelength, frequency and offset.
-emit from geo surface.
Once I had the first one setup, it was easy to dupe it and change the setting so that the 2 tendrils work in tandem.
This is what the revised ‘2 tendril’ component looks like in isolation –
Now that I have discussed a bit about how the tractor was created and the changes I made for the final, here is a bit about the workflow I adopted that worked well for me.
Using separate scenes for different simulations seemed to slow me down and I felt unorganized generally with this original method. So decided to have a crack at doing it all in one scene and that felt better for me. This here is the list of steps I used as I took each particular simulation from testing through to rendering-
– build everything required for the test, playback in the timeline to get a feel for whether its working using very low particle counts.
– group everything involved with that particle sim i.e curves, joints, deformers, emitters, particles, fields etc into a group with a specific name.
– nest this group into another group one up in the heirarchy called ‘tractor FX’
– as things start to look right, increase the particle count until the movement is perfect
– start playing with render attributes, shaders and render settings, doing single frame renders improving the render look – colour, opacity, size of particles, motion blur etc.
– when things are looking perfect – cache the particle to disk and render a frame range of about 30-40 interesting frames
– open the sequence in nuke, add it to the stack with an over merge node and see how it fits in
– go back to maya, change anything that needs tweaking and render all the frames for which this particle is involved.
– back to nuke – update the read node to include all the frames
– if happy, then delete the particle disk cache files (I have limited hard disk space), turn off ‘is dynamic’ and hit delete cache.
– hide that group in the outliner so that it no longer gets in the way
– move on to the next simulation and start from the top
This way of working was very productive for me, I am a control freak and like to be able to know exactly what is going on.
For safety, I did a lot of intermediate saves. I ended up at save number 23-
This snap shows how I organized the outliner–
RENDERS ATTRIBUTES/SETTINGS –
I thought it worth talking a bit about my render attributes and render settings for all the final renders.
I never thought I would end up using hardware instead of mental ray because I am a pretty big fan of MR usually.
What I liked about hardware is that it seemed to have a real ‘particly’ feel. In the testing I did with MR, everything looked just a little bit ‘cloudy’. Good for some things maybe but I preffered the look that hardware gave me. When I referenced it back to what I liked about Legend of the Guardians, it was this ‘hard’ look that suited what I was after.
I did a lot of testing for the final look but when I got it right, it meant that I could use virtually the same setting for all the particle sims. It was always multi-point and in fact, you could argue that I could have used one particle node and linked it to all of the emitters? The reason I didn’t was that I found myself tweaking things like ‘multi radius’ slightly differently for each one.
This was the ballpark setup for all of the particles –
– Depth Sort and Col Accum both on
-Multi-count of around 2
– point size – 1 or 2
– ramp for opacity PP – very low max values in the array mapper to keep the opacity really low overall. Somewhere between 0.3 and 0.05! This was crucial to the look – high opacities DESTROYED what I was trying to acheive!
– colour ramp on colour PP – from a light blue at birth to just slightly darker blue at death – again, darker death values really killed the effect!
– for the actual render settings, number of samples at either 1 or 9 depending on the detail required versus speed that I needed.
– no motion blur – I would like to experiment with this more in the future. I don’t know if I was using the settings properly but when I used any motion blur it just looked too soft? Using slightly lower ‘mix’ settings in nuke and subtle nuke glows seemed to better help the look I was going for.
Other (non-tractor) FX –
Heat Plume –
I made some poly planes to emit blobbies for use with the iDistort node in Nuke. I had them in three separate banks – lower, middle and top because the doors do not all open at the same time. This way I could stagger the emission of heat plumes to match up the timing. I just animated the emission rates to begin as each door opened.
To the rag doll animated using dynamics, I applied a series of dynamic constraints to the separate pieces of geometry. My methodology basically went like this –
– for joints where a 90 degree rotation was required, i.e -the knee, apply a directional hinge constraint.
– for joints requiring a 360 rotation i.e the neck, apply a pin constraint.
– move the constraints pivot to the point of rotation.
– the key to not getting interpenetration problems was to skip the in-between joints- for instance – creating a constraint from the upper leg to the lower leg, skipping the knee.
– this caused the obvious problem that the knee would be ‘left behind’ when fields where applied.
– I solved this by parenting the knee to the upper leg in the heirarchy. Before doing this, I would move the geometries pivot point to the pivot of the parent so that they would rotate around the same axis point.
– same method for the rest of the ‘interconnecting’ joints like the neck, waist shoulder elbow and wrist.
– then to get the rag doll moving up towards the ship, I applied uniform fields in negative Y.
– to make the movement more ‘dynamic’. I only applied these uniform fields to the hands. One at magnitude 500 and the other at 1500. That way, the rest of the body would be ‘dragged’ with the arms leading the action.
– To add some more interest, I applied a vortex field to the chect to get a slight bit of oscilation.
– for the render, I created 4 directional lights all of varying intensities and shades of blue to create the reflecting light from the tractor.
– I had to scale down the blue saturation in the comp as it was too strong but at least it gave me something to work with for ‘integration’ into the scene.
– Since the camera was static rather than moving, I did not bother with holdout masks for disappearing into the ship, I chose to just make a roto shape in nuke. This would save me from re-rendering if the client changed the ships position –
Ground Fog –
I promised myself that as a last step, if I had time – I would try to create a ground fog effect using particle sprites smoke. I wanted to do this for a few reasons-
– it would give me a ‘canvas’ to bounce some blue light effects off in nuke when the tractor beam comes into play.
– I thought that the foreground of the shot looked a little dark and led the eye away from the ‘hero’ tractor beam. Blending/lightening/smoothing this area with some fog might help the composition.
– The first 300 frames are pretty boring with not that much to look at – some moving fog would add some action to this part of the shot.
– some fog might help to ‘unite’ all the elements in the scene – if the fog moves faster as the ship approaches then it could help feel the drag/displacement of air from the UFO?
How I did it?
I created one single cloud texture by painting a single stroke on a paint FX canvas with a cloud brush, saved this as an iff file.
Emit sprites from a poly plane just below the visible crop in the scene.
A variety of runtime and creation expressions to randomize the size, twist, opacity etc as follows –
runtime ex –
smokeParticleShape.spriteScaleYPP = smokeParticleShape.spriteScaleXPP = smokeParticleShape.spriteRampScalePP * smokeParticleShape.spriteRandScalePP;
smokeParticleShape.spriteTwistPP += smokeParticleShape.spriteRandTwistPP;
smokeParticleShape.opacityPP = smokeParticleShape.spriteOpacityRampPP * smokeParticleShape.spriteOpacityRandPP;
Creation ex –
smokeParticleShape.spriteRandScalePP = rand(4,16);
smokeParticleShape.spriteRandTwistPP = rand (-1,1);
smokeParticleShape.spriteOpacityRandPP = rand(.1,.4);
Then I found a point in the sim where there was a good amount of fog and set initial state so that they would already be in the scene at frame 1.
The the plane stops emitting and the fog gets shifted around with turbulance/ unifom fields as the UFO arrives/ leaves.
Here is a couple of seconds of what it looks like on its own-
I had to mix this way down in the comp as it was a bit distracting but I think it works. It also had some artifacts from the sprite shapes so I had to roto some blur nodes in to fix this –
Final Comp result –
I had fun on this project. It had so many elements that at some points it felt a little overwhelming. It forced me to adhere to my own naming conventions, think about workflow, testing and getting one thing right before moving on to the next.
As far as the end result, I still don’t think the tendrils match the vision I had in my mind but I am pretty happy with them. As far as everything else, I am wrapped. I was able to make the changes asked for and to work to someone else’s instruction/vision which is a good indicator of how the progress has come along.
Things I would do differently if I did it again?
– I would have persisted with mental ray further and see how I could change the look of the particles.
– I would investigate motion blur settings a bit more and see if they could have improved the feel.
– I would experiment more with ‘particles emitting’ particles a bit more to closer match my reference material with regard to ‘trail’ effects.
– I would try other solutions to integrate the tendrils with the rag-doll. To combine a solution that actually wraps the particles around the arm for instance.