This term I will be focusing on animating my model.
Firstly, I will re-do the displacement map, to see if I can add any improvements where I am not happy with the character.
The first thing will be to create a human rig for the model. Following tutorials from YouTube I will be able to do this using the skeleton and joint function when rigging in Maya.
Next I will need to connect the skin with the joints then paint skin weights.
If possible I would want to rig the face, it would not need to be too complex, and I am contemplating using blendshapes only.
Animating:
Once the body is rigged and working, I am planning to animate a walk cycle on my character.
I have done walk-cycles before on Maya, so I will re-watch tutorials regarding this.
I will also want to animate the clothes on Marvelous Designer, as well as the ensure the hair moves correctly in Xgen. I will watch tutorials regarding this.
The Environment:
For the environment I have decided to go with a natural landscape, involving greenery and flowers etc. I am planning to take some pictures for inspiration, and following two artists’ work that I like.
Inspiration / Moodboard:
To Do:
- Re-do displacement map (focus around eyes, mouth and knees)
- Create correct hierarchy of joints for skeletal rig of model
- Connect skin and joints – paint in the skin weights
- Use Blendshapes to animate face slightly (Blink, mouth, cheeks etc.)
- Animate walk cycle (how to animate clothes and hair?)
- Start creating the environment (possibly using procedural effects?) using reference images of natural landscapes
Firstly I went back to my Zbrush model and re-sculpted the details, following better reference material.
I was working on closing the holes in the mesh for example within the mouth – it would make my UV’s better and more accurate. I’m not planning to animate the mouth so this won’t be an issue later.
At this point I imported my high-poly model into my Maya scene, as I want to see how the details render, if the proportions are accurate and if it will fit seamlessly with my existing hair and eyes.
I was going back and forth a lot with my Zbrush model and Maya, trying to get a realistic sculpt and testing out the renders.
As I want to animate this model, I needed to use the low-poly version with a displacement and normal map to show the details in the Arnold render. Again, this involved a lot of back and forth with Maya and Zbrush.
Getting the right displacement settings in Maya was very difficult, I was struggling with understanding this fully and was relying on various YouTube tutorials to figure out what would work best with my character. I found that displacement settings are very subjective to your own work – I couldn’t just directly follow one tutorial.
However, eventually I found a great tutorial that involved using aiMixer and aiLevels in the Hypershade with my displacement shader.
I followed a tutorial on creating human skin in Substance Painter. Following what I learned last term for the placement of underlying skin colour and inconsistencies.
I exported the base colour map to my Maya scene, and played around with the displacement and bump depth.
I’m happy with some of the detail on the face, but the skin needs more work.
I have an issue with the face shape. At first I redid the original model in Zbrush, using the sculpt and move tools to create the look I wanted. Then I re-exported the displacement and normal maps and input them all into my Maya scene.
This did not work well.
Instead I decided to manipulate the face shape in Maya.
I used a mix of editing the vertices with a soft brush and the sculpt tools to create a shape I liked. The chin and jawline needed to be smaller and much more defined in my opinion.
I also re-did the eyebrows and eyelashes in XGen, using the same base as my previous model meant that the process what a lot faster. I used vertices to move the base to fit the shape of this model, then refreshed the Xgen splines that I had already created. I moved them lower on the face. Its too heavy at the moment, so I’ll work in the density next.
My hard drive became corrupted. This meant that I could no longer access any of my files.
I managed to recover a majority of them using a software online.
I then needed to sculpt more details on the rest of the body. Firstly, using my own hand as reference I added the detail to my characters own.
Once done, I imported to Maya to see how it renders. The base colour isn’t accurate, I plan to add more variation and subsurface details later.
At this point I am happy with the head and hair.
Next I needed to start rigging my model. I watched a YouTube tutorial regarding this, we had also learnt the basics about creating Joints and Hierarchy in our first term.
I started by using the Create Joint tool, and placed the joints in the correct positions and order as needed for a human skeleton. Going between viewports to ensure the positions were within the body.
Following the tutorial I was watching I input the correct data into the script editor to ensure the joint positions were accurate for a human model. Once done, I selected the rig and my mesh and did Bind Skin.
I used the connection editor to create a driven key between the different joints and used NURBS curves to create handles for my rig, connecting them via parent controls. Throughout this process I ensured that each object was labelled correctly in the outliner, as the naming is imperative to organisation of the rig.
At this point, the rig was working, I needed to paint the skin weights according to the influence of each joint. This was quite difficult at first, once I spoke with my tutor he advised me how to use the skin weights brushes in an optimal way. He suggested using the Add tool only, not Replace.
He also showed me how to move skinned joints, as we realised the clavicle and shoulder positions weren’t correct.
Once I had finished painting the skin weights I tested out poses for my model, and how they render.
Furthermore, my tutor and I decided to use blendshapes to manipulate the finer details of my mesh. For example, the shape of the armpits and shoulders, we also put a set driven key on the animation – when her arms are raised the armpit blendshape will be activated.
At this point I needed to move onto my environment. At first I was advised to look into downloading assets for my scene. I looked into this but what I envisioned was too expensive. I decided to model it myself as was the initial plan.
The first thing I did was create a plane in Maya and roughly manipulate the shape to what I imagined for my scene. Using the move and scale tools on the faces and edges of the mesh.
Once happy with the size and overall shape I imported the model to Zbrush to sculpt in the detail. Primarily using the Standard and DamStandard brushes I sculpted mountains according to various reference photos for inspiration.
At this point I imported the .obj to Maya to see how it renders with the water plane in place. I just added an AiStandardSurface with the water preset at this point.
I did a few renders to see how the skin and model are looking, I also wanted to test out the dress from last term on my model.
I wanted to redo parts of the Xgen hair, in particular creating a parting down the middle. Again I watched a YouTube tutorial, it involved creating a mask on the head dupe mesh (which the Xgen guides are placed on) and painting either red or yellow on each side, so the guides know where to create the hair.
Again at this point I wanted to test out how the model moves and can be positioned, furthermore how this actually looks in the render.
Once happy, I went back to my Zbrush model of the environment, and used the paint tools to add colour as needed, which I could then export as a colourmap.
I also decided to use a displacement map for my mountains, so that the polygon count of my scene would not be too high. So I exported it at the lowest level and added displacement and aiNoise to create the detail and contours.
I found tutorials I would need for the later part of this project.
Now I needed to work on animating my model. After a lot of back and forth with the position of the skin joints and painting skin weights I started to pose and animate my model.
At first I blocked out the basic shapes and forms, using key frames.
Initially the idea was to create an animation of the character flicking hair out of her face. Very quickly I realised this would need more work and finessing than I had time for.
Then I decided to have her putting both hands on her hips instead, it seemed more achievable in the time frame. Next, starting to work on creating and animating cameras for my final shot.
I used the graph editor throughout this process. Ensuring the hands and arms were positioned and moving correctly and smoothly. I wanted the animation to look like slow-motion.
When animating the fingers, I needed to re-do the skin weights on the joint influences, I also realised there was a joint missing that needed to be added just before the fingertip.
At this point the animation was roughly in the right place. The idea is that the character is acting ‘idle’, so the cameras can pan over the clothes and face.
Now I needed to create the clothes. Last term I made a dress in Marvelous Designer, so I remembered vaguely how to use the software. Once again following a tutorial online, I created a ‘fitting suit’ of my imported character, which means the software can understand important points on the figures body. Therefore the clothes can be placed correctly.
At this point I came across an issue.
When working with MD I soon realised that scale is very important, as it works at human size. My Maya scene however was not at an accurate scale. It was almost 3x smaller than it should be. By this point I also couldn’t re-scale my whole scene, so I decided to re-scale my model when I imported into MD. Creating the fitting suit ensured the software could understand the proportions in relation to the clothes even after re-scaling.
I started by creating the pattern for the jeans. I haven’t made jeans from scratch before, but I’ve seen the pattern so I knew roughly the shapes needed.
After making the jeans the fit I wanted, I applied a denim texture to the fabric so it behaves with the correct properties.
I had already made a corset for my last term, I used the same pattern but again because of the scaling issue I had to re-do a lot of it. Particularly the back lacing had to be re-done, I watched the tutorial I used last time to remember exactly how to do it.
Once made I organised the patterns so that they would be imported as clean UV’s. I imported the model to Maya, and applied the denim texture I used on the dress.
I also at this point started working on the animation of the clothes, I had the animation of her body done, the only thing left was the fingers. So I exported the animated body as an Alembic file into my MD file, then reset the clothes on her body and played the animation.
Marvelous Designer automatically simulates the clothes according to the body animation.
Once happy with it I exported the clothes as an Alembic cache to be put back into my Maya scene. As stated earlier, the clothes were 3x too big so once back into Maya I just changed the scale to 0.333 and it fit my character.
I also started exploring how I would create the greenery in the scene.
I used the paint effects in Maya initially, however they don’t show up in the Arnold render view unless converted to polygons. This then needed to be applied with an aiStandardSurface. I realised this technique would not work as it was way too heavy within the scene.
At this point though I like the overall composition, again creating renders to experiment with the look and camera angles.
So the next option was to try to use MASH to distribute the greenery. Following a video on YouTube I learnt how to correctly set up a MASH system.
Firstly I needed to extract meshes from the original mountain, specifically in the area I want the grass. Then I would use Maya’s paint effects to create a small patch of grass on them.
Next converting these to polygons, and adding a ramp shader.
This grass then needed to be exported as an Arnold Standin (.ASS). With the Standin re-imported and selected I created a MASH network and therefore a MASH instanter. For distribution type I chose mesh and input the plane I had extracted for the grass earlier. Playing around with the density and randomness I covered the surface with greenery.
I did the same thing for the trees.
Then I wanted to explore creating flowers, again following the same steps I used for the grass. I found that with the flowers they took much more time to achieve the look I was after, I needed to apply different colours and textures to each individual part of the mesh.
After experimenting with what greenery was available in the Maya content browser, I found my scene was again getting quite heavy. I lowered the density of everything. At this point however, I am liking the overall render.
I’m starting to render different angles to figure out what frames work for my final animation.
At this point I wanted to re-do the water and animate it if possible. I followed a video, where I added a noise node to the water plane, and then played with the settings and depth to create the look of waves. Then keying the time section of the noise node so that there would be movement imitating water.
I also added slight transmission and a blue specular colour.
At this point I was starting to finalise my camera animations.
Next I followed a tutorial on how to connect the eyelashes to the eyelids, then I could proceed with making the blendshapes for the blink. I needed to create a set of joints across the edge where the eyelashes meet the eyelids and use a parent constrain to connect them.
Then I started animating a blink, I also wanted to create a smile, following references online I started working on this. We learnt how to use blend shaped for facial animation in term 1, so I referred back to those videos.
Once happy with the blendshapes, I started working on the detail of my original body animation. Again, following references I found online I tried to get accurate movement, using the graph editor and my rig controls.
Then I started working on the overall details missing on my character. I started by extracting the UV’s on my clothes mesh, so that I could texture each part accurately. I used the denim texture from last term, but played around with the scale of it in Photoshop, tiling the image so the size of the denim fit the UV’s.
I also used an image of the Dickies logo to put on the pocket.
Then I decided my character needed nails, I created a very simple shape of a nail from a polygon primitive cube.
Placing them in the correct position and scale for each nail, I then added a ramp colour. I parented each nail to the fingertip joint in my rig so they would follow the animation.
The nails could definitely do with more work, but at this point I knew what would be in frame of my final animation, so this level of detail would suffice.
Now I needed to create the final part of the texture on my models clothes. It is based on an outfit that I have already made.
I started by roughly drawing out the positioning of the dragon in substance painter, I did this last term so I knew the method would work. In substance painter drawing on the 3D object corresponds to the correct UV position.
I imported the new texture with the line drawing into my Maya scene to see if it matched up between corset and jean. This required a bit of back and forth between the two software.
Now that I knew the positioning, I used Procreate on my Ipad to sketch in the colour and detail on the UV’s.
Again, I started to finalise my animated cameras, using the graph editor to ensure the movement was smooth.
Now I wanted to attempt to animate the Xgen hair, this was quite ambitious as I don’t know if the scene can take it.
Again, I watched a few YouTube videos, firstly I needed to convert the Xgen guides to NURBS curves, and attach them to the hair (using the modifier AnimWire) and put it in colour view. These then needed to be animated using modifiers on the ‘HairSystem’ node. I started with the Dynamic Properties section, changing the drag and mass, as well as the stretch and compression resistance. I did all of this following a tutorial.
Next I changed the settings on the ‘Nucleus’ node, changing the solver attributes and adding wind speed and noise.
Once animated I needed to select all the NURBS curves and create an Alembic cache of the animation data. This would make it easier to play back the animation. In the AnimWire modifier I turned off live mode and referenced in the Alembic.
For some reason, this didn’t actually make the playback of the hair any faster or efficient. I did not have time to figure out why, but I could see that there was an animation on it.
At this point I was ready to render (I also had to because of time). I had created the animated cameras, my sequence was 1000 frames long, after testing out one frame at the correct render settings it took roughly 8 minutes. This would not render in time, with help from the technicians we tried to use the render farm but there was an issue with the file paths.
The solution was to use 5 external hard-drives and render out the sequence in series of 200 frames.
So there was an issue with the animated hair and water. In my final render sequence I realised the animation I had applied to the hair was not playing. I didn’t have time to re-render for the showcase, however I managed to get it working for my animated poster. I also added a free water and birds sound to my animation, fading it in and out in alignment with the visuals.
The water I will re-animate at a later date. I also had to re-do a lot of the MASH greenery last minute, so with time constraints the flowers had to be cut.
Upon reflection of this term I think I was very ambitious with what I wanted to achieve, I am overall happy with my output but there is still a lot of elements of the final animation that I’d have liked to fix. Perhaps focusing more on the animation and finessing the details rather than building the environment also, or vice versa.
Regardless, I feel I have learnt a alot about how to create an animated character from scratch, which was my initial desire from this project.
Another thing I learnt was to leave enough time for technical problems with rendering before a deadline. Ideally I’d have liked to render my sequence a few times so that I could fix any issues.