#assessment 02

After some extra feedback from peers, I’ve decided that the hair still isn’t relatively convincing enough from where I’ve left off.

So I decided to update it further by ‘flattening’ parts of the pointy bits on the side to make it appear more defined, as well as making them drape down more so that it appears like they’re being affected by gravity.

My decision to flatten the areas around the ‘sides’ was loosely based on my observations from a sculpted character bust that me and Sorin found on Pinterest (see below); sculpted by Adam Esat on ArtStation.

Here is my rendered turntable of the final version of my Honeydoo model, including a wireframe render. (Use this link if it doesn’t appear above: https://youtu.be/DOnKmIxNraQ)

I used the PDFs provided by Thomas in his Advanced Animation Production, composition workshops to help me get it set up. Although I was unable to get render layers working properly, which is something he too had trouble with.

I’ve made adjustments to the mesh, especially around the hair in order to make it appear less like a helmet. I’ve attempted this by making her ears slightly visible, making the overall shape of the hair less ‘perfectly round’ as well as basing the side profile on that of actual afro hairstyles (see reference far below).

I’ve also updated her hands so that there’s an extra loop at where the fingers begin and I’ve moved the thumb down to a more natural position. 

Ideally, I reckon they could use a lot more edge loops for extra definition, however since this is smoothed version (for sculpting) from the low-poly model, it might be best if I just did this by revising my low poly model with what I’ve learned from my first sculpt-modified version.

I’ve also made attempts to create an updated version of the model with additional detail, especially on the shoes. The arms have also been slightly updated to appear less sausage-like.

Here is my finalised version of Honeydoo on Sketchfab.

As a bonus, I also did a quick deformation test of one of the points on Honey’s mane. The pivots aren’t ideal, probably because I didn’t add enough joints, so it’s really easy to break.

Here is a list of blendshapes I produced for Honeydoo to test her facial deformation. They’re based on the 7 universal expressions, as opposed to the modular facial system I had for Perry. They’re kind of based on the diagram shown on the lower image.

I gave her a ‘default exp(ression),’ which closes her mouth, which I had originally modelled open.

Here is a video demonstrating the blendshapes in action.

I was originally hoping to just animate and playblast them like I did with Perry. Unfortunately, the blendshapes seem to glitch out when animated, resulting in the shading changing in a way that makes Honey appear uglier and uncannier (see ‘laugh lines’ and cheeks below), as well as the very tip of the nose being unable to move.

Sorin would give me additional feedback for the expressions by reminding me to make the jaw movement more obvious; keeping the upper teeth still and pushing the shape of the eyes further.

Sorin also mentioned that the eyelashes and their relation to the eyes tended to appear blocky. He pointed out that it was likely because of my decision to make the eyeliner a seperate material (shared by the lashes), as opposed to sharing it with the skin.

As such, I updated the model so that the entire face mesh uses one material. Below are some renders of the seven universal expressions with my corresponding update.

When working on the renders, I noticed some issues involving the geometry when that I did not notice before, without lighting, such as shadow artifacts on the chin and odd rims around the lip.

I also wanted to point out that the hair materials seems to be considerably darker and less saturated than the mane for some reason, despite sharing the same material. It later turns out that the wrong hair material was used and there’s somehow two of them in my Maya file.

Below is a collection of higher quality stills of each expression.

After taking into Andrew’s advice on always enabling tessellation and having gentler use of the smooth and bulge tools; as well as being gentler with the grab tool in general. I was finally able to sculpt a mane that (although resembled some kind of weird starfish, sea creature) I was pleased with because it stayed true to my reference sheet.

So I brought my sculpted model, at the lowest division setting, from Mudbox back to Maya to retopologise the mane and make and final tweaks before sending it to Substance Painter for texturing.

Here is the very start of my retopology process, using Maya’s quad draw.

As you can see later into my sculpt, I hid the body mesh so it didn’t get in my way. I would later hide more and more of the meshes, such as the hair and the teeth as I worked my way into the centre.

The points/fluffs were going to be tricky, especially in regards to handling the topology flow and their relationship with one another across the mesh.

As you can see in these two images, I began running into major issues with the centre edges merging into each other and Maya’s symmetry feature becoming unreliable. This is when I attempt to relax the edges along the geometry or add additional edge loops near the the centre-most edge.

Note the faceless edge at the centre of where I deleted them in order to fix them.

Here is the topology as it appears again, with the hole filled in, smoothed. When unsmoothed, the mesh appears perfect, but it is revealed there are overlapping vertices.

Here is a video showcasing the symmetry ‘merging’ issue in action. 

At one stage I managed to work around the issue (see lower image) by temporarily disabling symmetry, relaxing the area a bit and then re-enabling it, followed by mirroring the mesh so it’s truly symmetrical again.

Unfortunately, as I continued working on the topology, that’s when what’s depicted in the video started occuring again.

Here is the topology continuing through to the back of the mane.

Here is it progressing into the rear underside. Note the two bulges I accidentally left from sculpting. These would not be reflected within the retopology.

Here is my progress towards retoplogising the ends of the points. As shown, I began facing difficulties involving triangles and n-gons when trying to solve them. 

The usual solutions were to add more edges (which can make the model more resource-intensive and harder to work with) and to redirect some of the edges to make more of a proper loop.

Here’s an example of the topology on one of the fluffs, near Honey’s neck, showcasing a half-loop.

Here is my first completed version of the toplogy. I wasn’t pleased with how one of the points didn’t take its full shape properly, so I went back to retopologise it. Mainly with the addition of more loops and edges.

As you can see the geometry is very tight after adjusting the topology. Potentially more so than what’s preferable, but it’ll probably have to do.

Here is an overall view of the mesh, with a smoothed version of the retopologised mane, before I export her into Substance for texturing. 

Initially I was planning on modelling Honey’s ears/antennas in Mudbox, via tessellation sculpting. 

However I decided that it was too much unnecessary effort, since they weren’t supposed to appear particularly detailed and the quad-draw process would undo most of the work I did with retopology earlier. I’ve modelled the antennas with the ends straight, like in my reference sheet so they can be curved when rigged.

Sorin made another comment about her eyebrows being too lacking dynamics, with the uniform density. Just like he did with Perrys, which makes sense, since they’re copied and pasted from him without change. 

I’ve modelled a relatively simple ear cavity, which looks a little odd with the lighting/shading, since some of the edges are probably too sharp and tight than what they should be. Thankfully, it’s mostly hidden for the sake of this assessment and I can smooth it during sculpting.

I’ve made her ear appear ‘thick’ in order to make it appear more like it can support the weight of the antennas.

I used images of Team Fortress 2 characters (lower images) as reference to help me define the form of the cavity. Originally I had the cavity ‘point towards’ the face, to be more realistic. However, the shading looked more unusual that way hence why I changed it to something simpler.

Here is the overall view of the mesh before I put her into Mudbox.

My export process is the same as the previous post, smooth, duplicate then export. However, with Andrew’s help, we confirmed that the cause of the ‘twitching tool issue’ was from conflicting ‘contact’ from exporting the mesh with separate parts. 

The alternative solution, without combining the mesh, was to use the lock feature on the parts that I wasn’t working on whenever I’m sculpting.

Here is the updated UV with the antennas and ears included.

Here is the an overall view of my first sculpting pass, in progress.

 I mainly focused on getting the general shape of my character right, with most of the tweaks at this stage being in the hair (so it no longer looks like a shoddy helmet), the fingers and the chest (so she looks more like she’s wearing a shirt). I’ve also tweaked the shoes a bit, so the proportions are a bit more balanced and convincing; using references such as the lower image.

As you can see, the mirror functionality on the tools I used, failed to maintain its consistency on the other side shown by the grotesquely broken and stretched fingers on her right.

I plan on using Mirror in Maya to fix this, since Mudbox’s mirror feature is restrictive and would require more effort to clean up.

I also began attempting to sculpt her mane, using the tessellation feature, however my first pass turned out messy. Since not only is it poorly skewed but is also appearing too detailed and thus inconsistent with the rest of the character.

I also ran into some pretty bad ‘pinching’ issues with the ends of the points on her mane, where there’d be too much densely clustered definition in such a tiny area.

Smoothing it usually made the pinch tinier and worse, while bulging tended to ruin the established shape of the point and smoothing over it sometimes resulted in it reverting back to an unwanted pinch. So it often turned into an annoying, counterintuitive tug-of-war process.

Continuing from where I left off, I’ve begun to model the base meshes for the hair, the scarf/mane and the wings, as seen above.

Here is my solution for the n-gons inside Honey’s mane. It’s a bit stretched at the moment, but I was thinking I could get away with it, since I’ll be refining it during sculpting in Mudbox and redrawing the topology as required.

However I’ve later decided that the base mesh needed a whole lot more work done on it before I moved onto the sculpting phase; especially on the face, which I feel, and was recommended, needed to be brought up to the same standard as the Perry model. Although I originally had that in mind with the sculpting phase.

So I decided to add eyes and revise the topology of Honey’s face, based on what I learned from modelling Perry the Rabbit.

Here is a smoothed (preview), close up of Honey’s eyes, showing the work I did for the eyelids. The upper eyelid fold is loosely based on a still reference I acquired from a YouTube video tutorial series (lower image) about character modelling, when I was searching for “Pixar Eyelids.”

Honey’s eyelids are far more pronounced than Perry’s and they look more obviously Caucasoid, while I felt Perry’s appeared more racially ambiguous, despite my intentions to make both characters primarily Caucasoid.

Work needs to be done to ensure that the eyelids have a uniform slope along the vertices, otherwise if they’re wonky, they’ll look unintentionally wrinkled.

Here’s another view of Honey’s face, showing the eyelids more from the side, as well as further updates to the topology, especially around the mouth-nose area and the side/back of the head, so that it flows more intuitively. I’ve also added some additional edge loops leading to the bottom of the eye in the process, which will require moving around in order to preserve that round eye shape.

I’ve also decided to put some attention towards re-topologising her hair mesh, which is kind of like this weird, unappealing ‘helmet’ at the moment. This is to ensure the sculpting process goes a bit better with more intuitive edge flow, since I anticipated that all the unnecessary half-loops could complicate the process and limit the mesh’s flexibility.

Same applies for the mane-scarf mesh, which is a bit uneven at the moment.

Here are the back views of the same mesh.

I later modelled a mouth cavity, with altered copies of the teeth and tongue from my Perry model added.

Originally I had a whole lot more edge loops inside Honey’s lips, but I later removed most of them, bringing them down to the same amount as Perry’s, since they made mouth-posing a lot harder and more distorted.

I then modelled some eyelashes for Honey, based on a flat-plane, referencing the relatively simple, Grandma’s House Lucy eyelashes to get an idea of how to produce them. 

Clipping issues can easily arise from having them make close contact with the eyelids, especially when they don’t have the same amount of edge loops (which I might update) and when they don’t have the same colour as one another. 

So I began to colour parts of the mesh accordingly to get an idea of what Honey would look like without the plain old, grey Lambert; adding eyeliner with the inner-most eyelid faces in the process. Something that helps with the clipping issues on the eyelids a bit.

Here is an upper-body shot of the mesh with corresponding, single-colour materials and the updated hair/mane base meshes.

Same as above, except full body shot.

I’ve also updated Honey’s back topology a bit as well, deciding to add ‘elbow loops’ (circled in red), just like how I had with Perry later on (circled in blue).

Folds/circle loops for where the wings come out have been added as well, so they look more organic.

Ditto with the elbows above, except with the kneecaps. The extra geometry could be useful during rigging. I may need to add an extra loop or two, like I did with Perry’s in order to achieve more of a kneecap-appearance. 

Finally, I’ve decided that the base mesh needed a skirt to go along as well. Something that I was also kind of planning to do in Mudbox, but I ultimately decided that it wasn’t the right program for that kind of work.

Meanwhile, Andrew during this week’s tutorial recommended to solve the extra polygons on the thigh with a single triangle. As opposed to the half-loop that I had before, due to how unnecessarily odd he felt it was and because he felt that it was in a relatively obscure area.

It helps that I have the dress to also hide it as well.

Meanwhile, I’ve been thinking about how I was going to approach properly sculpting/modelling Honey’s mane. So I looked back at an older model which I’ve known for quite a while as a reference.

It’s a model of the anime character, Renamon, produced by KP-ShadowSquirrel; who has a similar mane, if not inspired that of Honeydoo’s. Upon inspection, it seems that ShadowSquirrel has modelled the mane in a way involving layers of closely overlapping, slightly modified duplicate ‘shards.’ 

Here is a mockup I made, showing what an improved version of the final outcome might look like, had I used 3D parallax and individual layers for each of the words.

It’s a bit rough around the lines and generally conforms to what I made previously. 

Although this animation itself isn’t really the best example, I learned that I can direct the flow/direction of the animation by using the ‘swing-door/blind animation’ in relation to where the next words/graphics are.

Looking back, I feel that the project could have been improved had it been put together with 3D parallax taken more into account. 

This could have been achieved by moving the (3D) text’s Z-axis; complemented by having each word in their own layers/composition, as opposed to using a series of animators in clustered, text layers.

This would’ve allowed more control over how the text was animated and used to compose the collage. Thus, allowing me to achieve more, dynamic visuals, reminiscent to Patrick Clair’s TED talks animation (2010).