Overall, the particle is driven by four elements. Two are quite simple, and Tim creates these forces in his tutorial:

- A “follow” velocity: The particle’s overall velocity direction is set to follow the tree curve. Before the curve is input to the particle solver, we can use the “PolyFrame” node to generate tangent vectors that point along the curve, then use this attribute to set the direction of our velocity.- A “suction” force: Force is added to the particle pointing towards the center of the curve, drawing it in. This keeps the particle from straying off-course and can create a slight taper in the the tree over time. The magnitude of the suction force is based on its distance to the curve.

Using Houdini for this particle simulation allowed me to repurpose complex particle nodes for unique needs. I added a POP Flock node to my particle system to take advantage of its avoidance algorithm, so any particles that get too close to each other will be pushed away, helping me resolve awkward wire intersections and particle convergence.

• Top of Page
• Home

One in a Krillion is a hack n' slash action game. As King Krilliam, control a massive swarm of krill to rise up against fish that seek to kill you! Your krill swarm can be used to form "constructs", giving you different abilities in battle, such as a stunning hammer, a protective shield, or a comboing scissor slash.This game is developed by a team of 10 people. I joined the team after one semester of development as a solo volunteer artist, and was responsible for the character models, environment art, and some technical art needs. This project was an opportunity for me to explore a simple, stylized, and cartoonishly goofy art style. Additionally, I contributed a handful of simple shaders to the project. Finally, working as a solo artist on a small project is quite an exercise in scope and efficiency!

Since this project is so free-form, I’ve been exploring several pipelines for asset creation, for the purpose of learning a new process. I created a simple rock generator in Houdini to help author individual rock props. This generator applies Worley (Voronoi / cellular) noise to a subdivided cube, with a large amplitude and element size, and low roughness. This distorts the cube into large cellular portions. The rock is remeshed and exported into low-poly and high-poly versions. I used Substance Painter to bake the mesh normals, and then create a simple texture that used my hand-drawn shape textures plus highlights on edges from a curvature mask.

As additional props in the level, I modeled several sand castle molds which act as walls and props along the edges of the level.

Basil and the Isles of Spice is a 3D plaformer game. Play as Basil, a basilisk, who must pass a series of challenges to become Shaman of her village!I joined Team Leviathan to work on this project in August of 2023. The project was in its final stage of production, with the goal of wrapping everything by the end of the semester in December. However, at the time that I joined, the game's level design had undergone a complete remake, which challenged our art team to rapidly create a brand-new environment.I created two tools in Houdini to help our environment artists with rapid asset creation and set dressing. These tools were integrated into Unity using the Houdini Engine plugin.Basil and the Isles of Spice will be on Steam soon!You can find my Houdini .hip and .hda files, as well as some basic documentation for tool usage, here.

One of my tools was a procedural platform tool, which creates simple platform geometry based upon a curve.This tool was initially tended for usage by our level designers, but it eventually evolved into an artist tool. Our game’s heavy stylized art style leaned into large and easily readable shapes, so the tool created geometry that exactly fit into our desired outcomes. By allowing artists to create platforms in-engine instead of having to use a separate 3D modeling package, we were able to save a lot of time, and iteration and changes could happen much more rapidly. While our key level structures were modeled by hand, most surrounding platforms were made using the tool.My topmost priority was to make this tool easy to use. The main controls are extremely simple: users could adjust the height of the platform and the bevel amount for both the top and bottom edges. A platform with a slight bevel could be used for rocky or grassy structures, while a very soft bevel along the top edge could create a sand bank, and a soft bottom edge could be used to round out a floating island.

Platforms in the Mangrove, Hub, and Arch areas, each utilizing beveling in different ways to achieve a unique look.

In order to make the tool both accessible and robust, I decided to automatically handle settings for clean mesh topology generation, while adding in some basic controls that allowed for situational changes.By default, the tool flattens the curve input to guarantee generation of flat surfaces. However, this can be toggled off if desired, to create sloped or sideways platforms.

Flattening the curve helps fix issues resulting from curves with y-level changes. However, toggling off curve flattening allows sideways meshes to be generated.

The mesh is automatically retopologized into evenly-sized triangles. This allows the mesh to be used with our grass tool in-game, which requires even subdivisions. By default, the entire geometry is remeshed; however, there is an option to only remesh the top and bottom faces, which can provide cleaner edges for less beveled platforms. The amount of detail preserved in remeshing can also be adjusted, so platforms which require less fidelity can be generated using fewer polygons.

Remeshing options can control the area targeted and the density of resulting polygons.

The Houdini node network for the platform generator.

The UI and parameters for the platform generator.

The Caldera, the ending area of our game. The bushes and kelp have been scattered using this tool.

The second tool I created for this project was a basic prop scattering tool. It accepts a ground mesh and a bounding object, and instances props across the mesh surface.This tool was initially extremely simple — it was more or less just a Scatter and a Copy and Align node, with Unity instancing of props and controls for prop size and density. I bashed this together during one lab period for artists to use in a few quick locations. However, as we found more potential uses for this tool, I began adding additional control parameters.One of the key changes was adding in noise to the density attribute. Simply scattering props evenly created too uniform of a look. Noise patterns, and controls for frequency, allowed us to create clumps of props, which provided a more natural look.

Coral and kelp scattered, broken up into clumps by noise. Both screenshots are the same scene; the right has water disabled for demonstration.

Iteration on this tool mostly focused on resolving performance issues. I quickly realized that, while a singular density attribute allowed artists to easily modify the look prop scattering, it also could easily lead to far too many props being added at once into the level. I added a parameter that capped the total number of objects scattered, and also double-checked the poly count of our prop meshes, and retopologized them as necessary. Finally, I added an attribute to automatically tag all instanced objects as Static, allowing them to be batch rendered and occluded.

The node network and parameters for the prop scattering tool.

As a deep dive into creating procedural tools for Unreal Engine, I designed a series of HDA’s to generate a city built upon Open Street Map data. This project was largely inspired by GDC talks: Marvel’s Spider Man, meet Houdini by David Santiago and Tools to Build a World in One Day by Thomas Tobin.You can download my final HDAs for this project and a demo final build here.The project builds a city based upon input of real-world map data from Open Street Map. The generated geometry adapts to terrain. Buildings are constructed out of modular wall pieces, which are instanced and can be swapped out for any asset in-engine. Similarly, instanced props can be scattered around the city, which can be used to place environment assets such as trees or rocks. A second type of prop generation creates props that follow roads, an ideal setup for creating lampposts, fences, or other similar props.

As a final step, I wanted to eliminate any buildings that intersected. To achieve this, as each building is generated, the network will check for intersections with all previously generated buildings. The new building will only be included in the final output if it does not intersect with any others.Given more time, I would’ve loved to add so many things to this network. A short list of examples include more complex roads, building foundations, optimization for intersection analysis (my current solution is very slow)! That said, the basic generation here is only the root of the overall project, and at this point I chose to move on in order to explore more tools for the system.

The “Building Generation” HDA is, to some degree, just a wrapper for the “Labs Building Generator” node. For its creation, I heavily relied upon Simon Verstraete’s tutorial on the Houdini Youtube channel — an invaluable resource. However, I had a fun time assembling additional logic to procedurally change the number of module & building variations.

The “Asset Scatter” HDA utilizes heightfield masks to control prop scattering. While I initially considered a manual setup for scattering points, Houdini’s heightfield nodes were more efficient and simpler to implement. They allowed me to utilize pre-existing features, such as the “Mask by Feature” node’s controls for masking by height and slope.The major feature of this network is the ability to avoid prop placement that overlaps previously generated buildings and props. However, the instanced nature of props in other HDAs present unique challenges when trying to avoid overlaps. I decided to use the “Mask by Object” node in order to delineate areas where props should not be placed, but this node can’t be used for packed / instanced geometry. (This is the opposite problem of the instancing issue in the building generator!) Like before, I fixed this issue by using the instance point clouds from input nodes. Cube geometry is generated on top of these points, then used to create a mask. Because of this method, the generated mask is not extremely accurate, but it’s acceptably reliable for preventing overlapping props.

Then, it’s a simple manner of using the “Heightfield Scatter” node to scatter points on the landscape. Props are instanced onto points.

By contrast, the “Road Props” HDA used a completely different approach for generating props. Like before, points are created, and then instanced props are copied to their locations. We also want to delete points that overlap other geometry. In this case, however, points are generated based upon the original road input.

The “Sweep” node, which we previously used to generate geometry for our roads, can alternatively be configured to make curves. Therefore, this can be used to produce two columns that follow the road, which then are resampled to set the distance between prop placement points. To clean up curves, I used yet another "Sweep" node to generate a bounding region that contains the roads. Any points that fall within this bounding region will intersect with the original roads, so they are deleted.

Like in the “Prop Scatter” node, other building and prop nodes in the network can be accepted as an input. Overlaps with these inputs will be avoided. I found myself running into the same issues I previously faced with instanced geometry. In this case, a similar solution — creating placeholder cube geometry in the location of input points — works again. This time, however, the geometry is converted into a volume, then used in a “Group” node as a bounding volume to select and delete intersecting points.

Beyond experimenting with various Houdini and Unreal features to create this project, I also learned a lot about managing my pipeline and workflow. Specifically, towards the end of the project, I struggled a lot with major issues with Unreal integration, many of which resulted from a lack of direction on my part.When I was developing the project, I often avoided bringing my tools into Unreal until the last minute. This led to many headaches where I couldn’t figure out why my network didn’t work the same in Unreal as it did in Houdini. In one particular example, I discovered that using a Houdini HDA as an input for another would only import the first output — I could not create HDAs with multiple outputs. Instead, I had to create groups for output geometries. These groups would be used to identify the desired data when working in another node.At the last minute, I experienced a plethora of build errors when trying to create a final playable build. At this point, it was nearly impossible to identify the source of the errors, and I ended up disabling several features of my project as a brute-force workaround. In reflection, I should have been making regular test builds throughout the creation of this project, to identify problems early and incorporate debugging into a part of my production pipeline.However, as a whole, I am proud of the work that I did on this project. Beyond the final product I produced, the development process of this project taught me a lot about problem-solving and procedural logic. After many weeks of work on these systems, I’m excited to bring the skills I developed to my next project!

• Top of Page
• Home

My first Houdini project: a tool that procedurally generates floating islands and bridges along a user-defined curve! In addition, the project is built so the user has a high degree of control over parameters for both the islands and the bridges.Making this project was a huge learning experience for me. The project involved a lot more Houdini VEX than I initially expected, and forced me to grapple scripting to a depth that I've never tackled before. However, I'm proud of the end result!For this project, I also produced a tutorial / explanation document on my creation process. The document works in tandem with my .hip file, which is thoroughly commented to explain what each node and script does. Download either of these below:

• Mirage can be found on Steam here.

• Mirage can be found on the Digipen Game Gallery here.

Jump through mystical desert ruins and shoot the bird-like guards to steal their treasure!Mirage is a short 2D platformer-shooter game, created as a student project by a team of 12 students over the course of a year using a custom engine.On Mirage, as with any small project, I held many hats. Primarily, I served as a tech artist: I created game VFX using a custom particle system, rigged our snake character and enemy characters using Spine, and implemented almost all of our game's art assets. In addition, I illustrated most of the background props and UI elements, and heavily contributed to style guide documentation and overseeing task management.The creation of this project was an invaluable experience in working with a team. Through both the ups and downs of this project, communication skills were a necessity for our success, and I often found myself serving as the bridge between artists, programmers, and designers, which was an incredible learning process for me!

This character is Alistair, a re-imagining of the character Alice from Alice in Wonderland. Despite being an old man, he preserves the original Alice's youth at heart. I imagined Alistair as an old man in the modern day, who still wears clothing that would've been fashionable in his youth - for this reason, I drew inspiration from a variety of time periods to give the clothing a vintage yet timeless feel.
At its core, this project was a study in stylization. Double Fine's Psychonauts (and other games, such as Broken Age) feature heavily stylized graphics, with characters built upon heavily exaggerated shapes. For this character design, I aimed to replicate their style, to gain a better understanding of the distinctive design decisions Double Fine employs in their media. This was much more challenging than I initially thought! I spent a lot of time iterating upon sketches before I started to find shapes and proportions that worked well for my design.

When I was in elementary school, my parents signed me up for a Chinese painting class during summer break. As far as I can remember, that was my first introduction to making art beyond crayon and marker scribbles; as I grew up, paint became my comfort medium to work in.These days, I enjoy painting as both a hobby and a way to train my eye and traditional art skills. Over time, I find it more and more refreshing to return to paint. It's a much-needed reminder of the foundations behind great art!All my paintings on this page are done using gouache.

Hello! I'm Annabel Sun, an artist currently based in the Seattle, Washington area. I'm a digital artist, specializing in 3D environment art & tech art for game development -- but I also have a passion for working with traditional media. This is my portfolio website: a showcase of my best work!I am currently a 4rd-year student at DigiPen Institute of Technology, studying for a Bachelor of Fine Arts degree in Digital Art and Animation.I am currently seeking jobs and internships!If you're interested in my work, please don't hesitate to reach out. I'd love to chat!
You can reach me at sunannabel@gmail.com.You can find all my social media & other links in the footer below.