3D Technical Director / 3D Technical Artist
manu3d @ gmail.com
Latest
Unrefined Design is a company creating decorative elements for interiors, made in a very fine type of concrete. One of their founders is a friend and is familiar with 3D work, as he created static renders of the products for their website. He asked if we could collaborate to create some animated content and this sequence is something I came up with, as an exploration on the topic. The work was mostly executed in Houdini, where I set up the scene, the lighting, the shaders and the water simulation. The rendering was then done with Redshift and the editing in Resolve. The bowl model has been supplied by Unrefined Design and the other bathroom accessories were purchased on TurboSquid. While I don't have a lot of experience with lighting and shading specifically, setting up the lights and the materials was fairly easy, just a matter of trying different things and eventually settling for a specific configuration. The water simulation, was certainly more challenging, including the fact that the presence of the water dramatically increased render time - the clips with water took one day for each second to render on my laptop and often generated GPU peak temperatures over 100 Celsius. Ultimately I'm reasonably happy with the results: some things could be better but I'm training myself to let go of -some- perfectionism and I'll be moving onto other projects and goals.
T-Cell 303 (Human Immune System) I take part in an informal, weekly "PsyPhi" (Psychology and Philosophy) discussion group and through it I met Fredrik Trulsson, a member of the Center for Scientific Computing (sciCORE) of the University of Basel. Almost by chance I asked him "Do you have any cool data you'd like to visualize?" and he said: "Actually... yes!" That data resulted in this visualization of an actual T-Cell of the human immune system. The team Fredrik is part of takes a collection of cells, freezes them in time by encasing them in a transparent resin block a few millimeter per side and out of it carefully creates hundreds of slices 40 nanometer thick with a microtome machine. The slices are then scanned with an electron scanning microscope resulting in hundreds of beautiful and insightful greyscale images that are used for further analysis. Nobody however had ever put the slices back together to create a three-dimensional object out of them, which is where I come in. Fredrik and his department kindly provided the data for one T-Cell in the form of folders containing hundreds of CSV files, one per slice, each in turn containing a two-dimensional array of the density values measured by the electron scanning microscope. Each folder contained only one type of component of the cell, for example only the Mitochondria, only the Plasma Membrane and so on, which was instrumental to assign different colors to each. In Houdini I wrote a little python snippet to ingest the data into volumetric objects with a resolution of approximately 1000x1000x200 voxels. Initially I used Redshift's volume shader to render all the components of the cell, but eventually I left only the Plasma Membrane (in dark yellow) and the Nucleus (in blue) as volumetric renderings. The other organelles, i.e. the orange Mitochondria, were converted from volumes to isosurfaces to polygonal surfaces and rendered with the standard Redshift shader. Besides the components already mentioned in this animation one can observe in bright yellow the Golgi Complex, in bright green the Endoplasmatic Reticulum Sheets, in semi-transparent green (reflecting their lower densities) the Endoplasmatic Reticulum Tubules and in red the Multivescicular Bodies. A Lipid Droplet, in magenta, is visible in right hemisphere of the cell. I loved working on this task, figuring out how to ingest this type of data and then finding ways to process it and visualize it effectively, displaying it with clarity without loosing its wonderful visual complexity. I am particularly struck by the intricacy of this cell in comparison to the typical illustrations of cells we see in school books. I'm glad Fredrik and his colleagues gave me this opportunity and I thank them for their support, enthusiasm and encouragement for the duration of the project.
Every Moment I was inspired by the "Sentences" work [1] of Belgian artist Fred Eerdekens, who creates amazingly beautiful curly 3D objects in metal and they turn out to be words when a light is cast on them from a specific angle. The words themselves in this case are from poet Walt Whitman and the William Narasi font is by Integritype Studio. The Houdini workflow I created for this piece is 80% procedural. Once the words and fonts are chosen I have to break the glyphs into curves (Houdini primitives) and provide a per-point "final_width" attribute. If I used just the outline of the font and a constant final_width it'd be simple: I'd only have to feed a group of cutting points to a PolyCut node, add the final_width attribute and I'd be done. For this piece however I recreated most of the brushstrokes of the glyphs as manually placed curves and an associated manually crafted ramp, indicating the width of the stroke along each curve's length - this was time-consuming. One constraint the workflow imposes is that each set of words has to have the exact same number of curves as the others, although corresponding curves do not need to have the same number of points. The workflow than takes care of the rest, including creating the trajectories between corresponding curves (via Hermite interpolation) and creating the animation. Initially I had some difficulties even thinking of the whole setup and its details as I wanted it to be able to handle a significant variety in terms of fonts, number of letters, sets of letters and timing - the system could potentially handle a much longer list of starting movie titles and ending credits. Working step by step through each little issue did the trick however, and eventually I had a breakthrough thinking of the animation as trains traveling with customizable departure and arrival times from one set of words to the next. Interestingly, under the hood the trains/strokes do not actually travel: a PolyWire node inflates the geometry to its final width but the underlying geometry actually moves very little and is always there, even if unseen, as a zero-width tube. On the rendering side I tried a number of materials and I loved the way a transparent material for the shadow-casting object generates beautiful caustic patterns and more structured shadows on the background - I might do a separate version for this. But I eventually settled for a golden material and increased contrast and desaturated slightly in comp for this published version. Thank you for watching! [1] fred-eerdekens.be/work/copper-works
Animated Untitled I was inspired by the work of Anna Kruhelska, who patiently creates this type of artwork in light resistant paper: https://www.annakruhelska.com/pulse-series/ I tried to capture the nuances of paper as much as possible, i.e. the way it scatters yellowish light in the interstices between the moving shapes. I also tried to match the original lighting setup and added some vignetting all around to help focus on the piece. It was fun to figure out how to create a flexible setup in Houdini, to easily allow for variations on the theme. In the process I realized Mrs. Kruhelska approximates mathematical functions with her sinusoidal lines but also adds some irregularities making her pieces subtly more interesting and personal.
MultiCube v038 Personal practice to solidify my understanding of Houdini's duplication techniques. I wanted a variety of cubes, each doing their own thing with random animation offsets, and the whole clip had to be loopable. This required a fair amount of VEX in Houdini, in particular due to the fact that as many of the cubes are dynamically carved by subtractive booleans, their point count changed which in turn badly affected motion blur. And as one cube's animation is finished, it is replaced by a different cube with a different animation and different point count. This means the output of the copy_to_points node is a single mesh with highly unstable point IDs. I therefore had to create a system before and after the copy_to_points node to provide stable point IDs from which to calculate the correct velocities for the motion blur. Very brain-tickling but very satisfying when it finally all worked. In the process I also had the idea to light up the interior of some of these cubes and initially I did it naively, with some emitting materials. This however resulted in grainy/noisy output that gave me the excuse to learn how to use a light mesh instead, resulting in smooth lighting and faster rendering times. This work was inspired by the series of (2D) Cube Icon Sets realized by "vecktor" and published on shutterstock. I initially wanted to do an animated version staying faithful to the original but it is not possible to contact the author via shutterstock to ask for permissions and I couldn't find him anywhere else on the internet. Also asking a shutterstock representative to clarify if I could legally create derivative work was not fruitful: the line between derivative and original work is not really sharp and I could not obtain a straight answer. As such, I took vecktor's work for inspiration but ultimately did my own visual brainstorming about what cubes to create and followed those sketches instead.
The World needs more Colors and Sparkles. Obviously. I was playing around with black and white rounded triangles to make an animated version of this pattern: https://www.pinterest.ch/pin/856317316658325394/ Didn't quite manage to do anything interesting with it and decided to crystal my way out of it. I know. Embarrassing. Moving on.
Carnivorous Plant AOVs This was one of my first ZBrush-based projects. I'm reasonably happy with the result and even happier that I now have a grasp of the ZBrush >> Substance >> Houdini >> Redshift pipeline. I feel that the topology could be significantly cleaner/simpler and there are all sorts of things that could be improved with the textures, but I really need to give this project a rest and move on. Also see: https://vimeo.com/888755999
PhoneBook 3D Model The turntable of a 3D model I made in the context of the CGMA course "Introduction to Production Modelling". The model was created and UV'd in Maya, textured in Substance Painter and rendered with Redshift via Houdini. For the course I originally rendered the same model directly in Maya + Arnold, but I wanted to check how close I could get to the same look with Houdini and Redshift instead. Turns out it was easier to obtain a nicer subsurface scattering effect in Arnold, but in Redshift it was easier to get displacement and normal maps to work correctly.
Templar Mouse A little character I made in ZBrush, textured with Substance Painter and then rendered in Houdini + Redshift. This is the first time I worked on a 3D character. The clothing was inspired by real-world reference photos. I was also interested in understanding the whole pipeline from ZBrush, through Substance Painter, until the rendering.