Personal project — 3D
The Pond Mage
A stylized creature beside a photoreal pond — a scene I made in Blender because the idea wouldn't leave me alone.
Image — Hero render
Drop in: the final render — the mage mid-spell in a moonlit field of poppies, a frog suspended in glowing motes. A young mage sneaks out to the pond at night to practice.
Role
Concept · character design · modeling, shading · lighting · environment · rendering
Tools
Blender (Cycles); also fluent in SolidWorks and Cinema 4D
Timeline
A few weeks · December 2024
Outcome
A finished night render: a stylized mage floating a frog at the edge of a photoreal pond, built end to end.
The bet
The idea
I wanted a stylized, charming creature living inside a photoreal world.
That combination is a known trap. Mix cartoon and real, and the eye starts prosecuting the seams. The character reads as a sticker on a photograph — unless the light, the materials, and the depth treat both citizens of the frame identically.
So the real bet wasn't "can I model a cute mage." It was: can I make the physics of the image so consistent that an impossible creature feels like he's standing in wet grass.
Every decision below serves that bet.
Decision
Water, at night
The setting bounced between a swamp, a river, and Lake Lagunita on campus. It just had to be water.
That was a technical choice first. Water reflects and refracts, and Cycles — Blender's ray-traced renderer — literally simulates light bouncing through a scene. If the pond behaved like real water, it would vouch for everything else in the frame.
Night was the other half. Darkness lets you choose what the viewer sees. The rig is two lights: a dim, wide moonlight across the field, and a warm, tight key picking the mage out of the dark. Cold world, warm subject. That's the whole mood.
The story fell out of the setting: a young mage sneaks out after dark to practice, lifting a frog gently off the ground.
Image — Lighting
Drop in: the two-light setup in the viewport. Cold world, warm subject — the mood is two lights.
Decision
Cute, not strange
I modeled the mage from nothing. The geometry wasn't the hard part; the reading was.
Cute is a razor-thin target. A head slightly too big turns creepy. Eyes slightly too wide turn alien. I nudged proportions for hours — booleans to cut the forms, sculpting to soften them, the loop tool to keep the topology clean. For tricky passes I followed a modeling time-lapse at 0.25x speed.
The wand got more love than it deserved. It's tiny, but it's the thing doing the magic. I stretched a cylinder, sculpted it down to a slim irregular stick, and unwrapped it for texturing — flattening its surface like a globe into a map.
Image — The mage + the wand
Drop in: modeling progression (blockout → booleans → sculpted softness), and the wand being shaped down from a cylinder. The smallest object in the frame got the most attention.
Materials
Shaders, explained like a recipe
A shader is the recipe that tells the renderer how a surface answers light. How much bounces off. How much soaks in. What color survives.
In Blender you build the recipe as nodes — little boxes wired together, each doing one job. One sets the base color. Another roughens the reflections. Another fakes tiny bumps without modeling them. Wire enough together and a gray lump becomes skin, or bark, or water.
Four recipes carry this scene. The skin sits on the stylized-but-lit-realistically line — believable under the warm key, simple enough to stay charming. The wand is matte and organic: rough, irregular, absorbing more light than it returns. The water mostly gets out of the way and lets the ray tracer perform. And the spell is an emission material — a shader that produces light instead of reflecting it. Each mote is its own tiny lamp. That's why the magic reads as energy, not paint.
Image — Shading
Drop in: the skin material mid-build, color picker open. Materials, built node by node — each box does one job; the wiring is the recipe.
Environment
One blade of grass, a whole field
Hand-placing thousands of grass blades was never going to happen.
Geometry nodes are the same node idea, but the graph builds geometry instead of surfaces. Take one grass model. Scatter copies across the ground. Randomize rotation and scale so nothing looks cloned. Expose density as a slider.
The result is a field I could art-direct live — thinner near the pond, thicker poppies in the midground. The pond itself is a trick I'm fond of: the ground secretly dips below the axis, and the water pools up to fill it. A real pond is just a hole that lost an argument with rain.
(Honest note: the grass model and the small background frog were sourced; everything else was built from scratch.)
Image — Geometry nodes
Drop in: the plane dipping for the pond; the node graph and density sliders. One blade of grass, scattered by a system I could tune live.
Physics
The experiment that didn't make the frame
The mage needed a garment. My first instinct was the honest one: don't fake fabric, simulate it.
Blender's physics engine includes a cloth simulator. You model a flat garment, pin it at the shoulders, and press play. Gravity pulls, the body pushes back, and the engine discovers folds no one would think to sculpt. Sliders decide whether it falls like denim or silk.
So I ran it. The simulated robe found beautiful, physically correct folds — and kept clipping straight through the mage's body. The fabric would settle, then a shoulder or knee would poke through the cloth mid-fold. Fighting the collisions ate hours.
It also taught me the deeper problem. A still render lives on its silhouette, and a simulation negotiates its silhouette with gravity, not the designer. I cut the cloth and went with a solid-body design for the mage instead. His outline became mine to control.
I'd make the same call again. And I'd run the experiment again — knowing from my own tests where simulation beats hand-work is worth more than either technique.
Image — Cloth experiment
Drop in: the simulated robe mid-settle, clipping at the shoulder, next to the final solid-body design. Physics found beautiful folds and the wrong silhouette. Cut.
Iteration
Then I let people tell me what they saw
Whenever I thought something was finished, I put it in front of people. I watched what they read — not what I meant. The gap was always the note.
The hat looked like a horn
Not one person read the pointed hat as a hat. A cone on a head has two readings, and anatomy beats costume. The fix was structural: a brim, the one shape a horn can't have. Silhouette is vocabulary. The brim is the word "hat."
The magic looked like a bug problem
My first spell — a particle cloud of glowing spheres — read as "a peculiar pillar of fireflies." The diagnosis: magic is a gesture, not a substance. A cloud is weather; motes streaming from a wand are a spell, because now the particles have an author. I repositioned the system and added the glow.
The frog was too cute
My adorable stylized frog didn't read as a frog. Worse, it broke my own rule: the scene gets one stylized citizen. I swapped in a realistic frog. Charm lost a round to legibility, correctly — and the mage got more magical by being the only impossible thing in the frame.
I could intend "hat" all day. If the room saw a horn, the design was wrong. The audience's perception is the spec.
Image — Iteration: hat + aura
Drop in: the "horn" version beside the brimmed rework; the firefly pillar beside the final wand-emanating spell. Nobody saw a hat, so I built a hat they'd see.
The render
Assembling the frame
The last layer is the camera. Depth of field keeps the mage crisp and melts the background. Our eyes associate focus falloff with real lenses, so the render feels photographed, not computed.
Under it: the reflective pond, the two-light rig, the geometry-node field, fireflies trailing into the dark. I came in fluent in SolidWorks and Cinema 4D, but character art, Cycles shading, and geometry nodes were new ground. I built most of it from scratch.
Image — Workspace
Drop in: the scene rendering on the cluster iMac. Where it actually got made.
Takeaways
What I took from it
The render holds: a stylized creature that genuinely sits inside a real-feeling world. That was the whole bet.
Three lessons traveled with me. Perception is the spec — the hat, the aura, and the frog were the same lesson in different costumes. Simulation is a collaborator, not a servant — the cloth engine found folds I'd never sculpt and a silhouette I couldn't accept, and knowing where that line sits is the skill. Realism is a system, not a style — the mage feels real because the light, the focus, and the water refuse to treat him differently from the world.
The tools change. That part doesn't.