Pre-Production
The Idea
The theme is ‘Adventure’, and taking inspiration from our classes on narrative our group put together a charming story about a cloud, on an adventure within himself to find his true shape.
This idea came from those stereotypical cloud watching scenes from films, and our own experiences of course! Looking up at clouds and trying to label their odd shapes as something recognisable, or ridiculous.
The Story
Our story follows a cloud character with a toony, stereotypical (kind of bland) shape. This cloud sees some other clouds playing in the distance, and he longs to be with them. However, looking at them compared to his own shape he feels inferior as he hasn’t got an interesting shape, and tries to copy their shapes exactly. This of course fails, as he has to find his true self (we imagined that clouds would probably take their shapes from things on the ground that they see and like! but no 2 clouds are the same). Saddened that he cannot form himself to look like the others, he floats to the ground and begins to rain (we also imagined clouds would rain when sad, thunder when angry, and create rainbows when happy!), this rain pools and beautiful flowers bloom; reminding the cloud of his worth. Now more confident, the cloud rises to the others.
Hero’s Journey
We tried to connect this story to the hero’s journey, the call to adventure being the journey to find a shape. The cloud’s challenges would be failing to copy others’ unique shapes, his revelation being that he is perfect as he is and mustn’t use others to find himself. His return is that he is happy with himself, and he floats up to be with the clouds he sees in beginning, now confident in himself.
The Inspiration
Though our story isn’t totally original, we thought it was a unique spin on the self worth narrative. We wanted our animation to target young audiences and be very toony, cell shaded and very bright with saturated colours. We took a look at other clouds in animation, particularly their styles.
Although this animation doesn’t relate narrative-wise to ours, we found it interesting anyway to see how 3D animation professionals make an anthropomorphic representation of cloud characters. Their models are very realistically fluffy and have other facial features aside from eyes and arms, which we planned on only adding (the less detail, the cuter we figured!).
https://www.youtube.com/watch?v=OybYKrIRK0Y
There were many 2D cloud styles we all really liked the look of, but translating them into 3D would be difficult. So we started looking more into how others made clouds using 3D software. There were impressively realistic renders, unique styles and cutesy bubbly clouds, we wanted to take a look at all.
We decided that very simple cute clouds would be the best, and started looking into how to make 3D scenes appear 2D with shaders, and edges.
We also took time researching real cloud formations, cumulus, cirrus, stratus, cumulonimbus etc. We assumed that some background clouds could be different variations, or we could have realistic positions like wispy cirrus clouds high in the sky, some low lying stratus clouds, and our main clouds would be bubbly, low cumulus clouds.
Storyboards and Designs
Everyone in the group had a chance to create their own character designs and storyboards, once done we could all collectively vote on designs and what parts of the storyboards we liked and would be added to the animatic.
All of our main characters looked very similar, as expected since he is a regular lumpy cloud! But we were more concerned about making unique and charming background clouds; mine were a rabbit with a carrot, ball and shoe, and a flower with a knife! In the end I modelled the rabbit (but with mini clouds, instead of a carrot) and a raspberry which Thomas designed (his design was originally a strawberry, but we decided on a raspberry for its rounder shapes).
I thought it would be funny, and cuter, if the cloud had few features; only toony eyes and possibly stubby arms. We decided arms would help us more with conveying expression, along with eyes. Some of our concept art had the cloud have a mouth, but for ease animating and cuteness, we worked only with eyes.
Another thing I thought would be fun is the behaviour of the eyes, as the character is a cloud (which doesn’t have a real body structure) the eyes would probably move around freely. Still I kept a flatter area in the middle, where the eyes would sit most of the time. Although the other clouds are drawn with eyes in the concept art, we assumed in the final animation that the clouds would be too far away and it wouldn’t be worth the bother.
Production
Making the Models
This video tutorial was perfect for our project, and was very easy to follow and recreate. Using metaballs for the clouds in blender was quick and, in my opinion, looks great.
Main Character
I wanted out main character to be soft and very rounded, to be very toony and cute, as well as easily recognisable as a cloud. The cloud is very smooth when in blender’s viewport but unfortunately it doesn’t stay the same when imported to Maya, it just slightly changes.
Because there isn’t really a facial structure, just an area where the eyes would mostly sit, and the metaballs aren’t too heavy in terms of geometry, I didn’t try to retopologize the model. Due to this character being a cloud, I had imagined in the character design phase that any facial features would move freely around the body.
These screenshots were of before I smoothed and flattened the ‘back’ side to instead be the front, so that eyes could sit on the flatter area.
Experimenting with influence settings
By default the influence was on 0.3, which looked quite nice but when experimenting with other number, we decided that the 0.1/0.2 range was the best for our model, as it matches stylistically with out toony concept. I adjusted the influence threshold in blender, on the first pass of the model to see how the influencer changes the mesh.
The Rabbit
Using my concept art as a reference I used the same method of building the mesh I used for the main cloud, using metaballs. I thought it was best to keep the rabbit on its own, with no carrot like in the concept as the mesh may have become too unrecognisable. Instead, I made some simple mini breakoff clouds to make the rabbit more interesting. This model took longer to make as its a bit more complex, but I’m happy with the overall look. As it is a cloud rabbit, I didn’t want it to be symmetrical or too obvious a shape, and I think I conveyed a good middle ground well!
The Raspberry
Surprisingly, the raspberry is the more simple design of the 2 background clouds and yet it gave me more trouble than the rabbit. My first attempt was messy and barely recognisable, I didn’t use any references or planning for this cloud and it definitely showed.
The original design was meant to be a strawberry, but a raspberry would be easier to translate into a cloud, due to the rounded shape and lumps. For the second pass, I made a plan on how to model the raspberry; start from the bottom with 1 small metaball, and work up with 4 slightly larger metaballs until the final layer is large enough to where they merge, and a simple stem can be made. I added the stem to make it appear more like a berry of some kind, I think it definitely helped.
When imported to Maya, unfortunately some metaballs from the main character model were missing and so I had to edit these back in blender. Because this metaball was added separately in the end, after the other metaballs had be combined, it does not connect as smoothly to the mesh and is quite noticable in this stage, even after smoothing the edges where the lump is in contact with the rest of the body mesh. Yet I figured with shaders and lighting it may not be a problem later on, so I kept the model.
For some slight variation in colour, I added lambert materials to the clouds with somewhat different, but natural looking cloud colours, a yellow hue for the main cloud, blue for the rabbit and purple for the raspberry.
Rigging the Model
Rigging was challenging for me, but I’m glad I attempted it and got the experience from this. I used Alec’s videos on making an IK and FK switch rig, the video was very easy to follow and the process was going smoothly, though taking forever, until I combined the rig with the mesh. I’m not sure what happened but certain parts of the IK and FK controls broke, I had asked the others in my group if they knew what was wrong, but we couldn’t understand it and having to go back and redo the whole thing might’ve taken too much time. In the end, I had to ask Alec for help with the rig. I plan on continuing to practise rigging when I can, to fully understand it.
Adding skin weights wasn’t as difficult as rigging thankfully, just time consuming. I gave adding the skin weights a go, but my laptop was having a hard time running after rigging, and froze, so I handed this job over to Thomas.
Backgrounds
Making the ground plane was quite fun and rewarding for me, as I could see how it was turning out with Arnold immediately. The ground was just a sculpted plane to seem like natural bumps and dips, and the grass was simply a clump from the content browser that was attached to the ground plane as an ai standin node. I ran into some issues with this later, as other members in my group couldn’t render the ground plane with the grass, so any other scenes with the ground plane would have to be rendered on my computer, but I didn’t know any other solution at the time.
I tried to make a bald spot in the ground, where I could later place a puddle. Using the video above, I added a small patch which later, and with changes to the grass, ended up being a much more natural looking dip. It was much easier than I anticipated!
After working on the ground plane, I decided that the plain brown colour didn’t look great when I fixed the grass to be shorter. So I found a simple grass texture online and applied this to the ground plane. There was an option for a second colour for the ground, so I just used a dark blueish green to create a speckled effect that I really like for the ground.
Lastly, I wasn’t sure at first how to make a growing puddle animation but then I realised that I could just simply use a short cylinder with high reflectivity, and edit the size to make it look like the rain is pooling. I really like the effect this gives, especially with the reflections. Its a bit more realistic looking compared to the toony style of the animation, but I think the pool looks very good anyway!
Animating
As a group we thought it would be best to keep the animations simple, and so I made in Maya basic eyes and arms. We wanted the eyes to appear 2D, and so I kept their meshes quite flat and very simple (I eventually rounded out the eyes, we thought they were too oval shaped). The arm in concept art was meant to be a thin, flexible tube with 3 round ‘fingers’, but I changed the design as having to model and rig those in time for our deadline wouldn’t have worked. Instead I used a spade shaped hand instead, something simple that later I would try to rig to be able to ball up, and I think this design works nicely. If I had more time and were to redo the models, I would definitely have tried making an arm and hand that matched our concept art.
The only animation aside from the blendshapes in my scene is pupil movement. The pupils move around the sclera (the looking at himself) and move slightly faster, before moving to the centre and wobbling a little bit to indicate sadness, that he is about to ‘cry’ (the rain) as he cannot change is shape.
Up close the pupils move a bit too much, but as the camera is zooming and and then is about to move down to follow the raindrops by this point I felt that having a more clear movement was best.
Blend Shapes
The blendshapes were tricky, I followed along with Alec’s video and got everything set up easily but I made the mistake of not setting up a driver. I just keyed the blendshapes in from the shape editor window and that led to problems with the motions; there were pauses and odd cuts. ( fullanim_gpcloud )
This method meant that the blendshape keys didn’t show up along the timeline which caused a lot of the movement and timing issues. I ended up having to reset all of the blendshapes and added the driver control, which helped a lot in seeing the motions on the timeline and being able to edit them in the dope sheet.
The eyes also got 2 blendshapes, one to appear sad and one for happy, but I used both for the last sad face shot! I thought having both, but coming in at different times, added more interesting movement and helped make a more emotive expression. I didn’t have to make a driver for them, as I had the animation already done.
Flower Blooming
This idea was given to us by the lecturers, to use content browser flowers for a blooming/growing effect, once the water pools in the ground. It took a while to get the flower curve to lay on the ground plane correctly, and even then the flowers were upside down! I had to flip it and try fix it, but even still many flowers aren’t sitting flush on the ground, and some are rotated awkwardly. However I didn’t think it was very noticeable amongst the grass, some are even better being so high above it, so I left the flowers as they were.
To keep with the toony theme, I just added ramp materials to the flowers in bright saturated colours, and I think they stand out just enough from the grass to be noticeable. Still, I had to shorten the grass quite a bit, and lengthen the flower meshes so that they were actually visible.
Making Rain
Going off of the video tutorial, I tried at first to have the nparticles coming from the cloud surface itself, however I quickly realised the particles would not only come from the base of the cloud, but from every part of the cloud! I only wanted rain to come out from the base, so I instead duplicated the puddle plane, changed it’s dimensions to mostly fit the bottom of the cloud, and used this surface as the emitter. Thankfully this worked, although I had to hide the plane inside the cloud mesh.
I also realised that streak shaped nparticles unfortunately can’t render in Arnold, so I attempted to work around this and use points but with a heavy motion blur attached, to make the particles seem less like rounded snow and more like lines of rain.
I had also fiddled about with the collision editor, trying to add a bounce, as when I rendered the scene the raindrops sat on top of the pool. I’m sure if I took some more time to fix this up the animation would have looked much better, but I was working late on a deadline! This is definitely something I would like to try fix one day, or at least redo it and figure out collision attributes some other time.
Toon Shaders
While searching for tutorials on making toony shading, to fit with out style, we found this very helpful video. I followed this video along, liked the results in the scene and went to render it with Arnold.
We were especially interested in having outlines, after Thomas showed our group what the cloud model looked like with lines for the whole mesh, we didn’t like it as much as we did the lineless versions. However as the cloud is white and has white eyes we instead decided to just make edges for the eyes only on a separate AI Toon shader.
I followed this video along, using a softer shaded look with 2 different shades of blue, and liked the results in the scene and went to render it with Arnold, to see what it would look like.
We quickly realised that this tutorial wasn’t meant for shading to be rendered with Arnold! I tried adding different lights, changing the interpolation, the colours, and specularity, etc; but unfortunately I just had to try a different shading method, as the technique in the video only resulted in my model turning purple.
Another issue I came across was trying to delete this material, I tried to delete it in the hypershade window but Thomas ended up helping and switched the materials.
Working Shaders
I finally got the shading to render with Arnold using this tutorial, I experimented a bit with this but had to restart. With help from Thomas in my group who was working on scene 1, we would all have to try replicate the shading for our own scene. Later, Thomas added shaders onto my model so that the scenes had the same shading settings, but I had to tweak them to match my lighting, instead of an aerial light I used just 1 directional light at an angle, to further match Thomas’s scene.
Finished Animation (Not Rendered)
It was a long and at times confusing process, but for my first full 3D animation with my own rigged models, I’m very happy with it! There isn’t as much body animation as I had done with the other assignment, aside from blendshapes and simple animations for the ground, but I’m still proud of the work. Making this animation showed just how complex 3D software is, and a lot of the time I didn’t particularly understand what I was doing or how to do something, and still I really did enjoy making it and I’m excited to learn more and practise when I can as this project helped me realise that I prefer 3D to 2D.
If I were to go back and change anything, it would be lack of animation in the arms and the grass nodes; the ground could only be rendered on my computer due to the grass, and I think some simple bobbing or hand clench could have made the animation more interesting.
Rendering Issues
My ground plane was to be used in Mary’s final scene, although when she tried to render it, the grass wouldn’t show up. None of us knew what was wrong, as it also didn’t work for Thomas, so I decided to render both mine and Mary’s scenes with my laptop, to allow the grass to show up. Rendering though took much longer than expected, it only seemed to render around 10 or so frames an hour!
Old Render Settings
I decided to stop the render around the 10 hour mark, as it had only got to frame 127 out of 450. The quality was very nice and everything looked clean, but I had to change my render settings so that the render time could be under 5 hours. This time, I changed the light samples down to 2, instead of 3 as they were before, and the camera samples down to 2 also, instead of 5. This took scenes with the cloud only around 18 seconds to render, and scenes with the flowers and grass about 1 minute.
New Render Settings
Final Sequence
I’m happy with the final render, I put the sequence together in premier pro. I think it came out quite clean with little noise and its very vibrant. However, if I could I would have changed the outline around the eyes, as when the camera pans down with the rain the outlines around the eyes thicken a lot for some reason. I’m not happy with the rain nparticles just sitting on top of the puddle, I could have tried dissolving these. Also, the flowers blooming is a little bit too fast and has an abrupt jump, which I wish I could have fixed. Overall, I’m happy and quite proud of my first ever full 3D animation, and I look forward to hopefully making others!
Final Animation
Once my scene was rendered, Thomas just had to render Mary’s as his computer would be the fastest. I put all 3 scenes together in Premier Pro, Thomas’ scene 1, my scene 3, and Mary’s scene 4 (as a group we decided to cut out Fionnuala’s animatic part, as it didn’t look right in with the full 3D animations and she couldn’t finish her 3D animation), and I edited the final animation to the music Mary picked. Even without scene 2 I believe the narrative stays relatively the same, but it would have made more sense if scene 2 was completed to show the cloud’s attempts at recreating the others’ shapes. Still, I feel for our first group 3D animation from scratch was a success!
Be First to Comment