Blog
What Is 3D Modeling?
3D modeling is the process of building a digital object in three dimensions, usually from points, edges and flat faces called polygons, so it can be viewed from any angle, animated or rendered. The four main ways to do it are polygon modeling, digital sculpting, curve and CAD modeling, and procedural modeling. A model for a CGI commercial has to be more than a good shape: it needs clean surfaces, correct scale and only as much detail as the camera will ever see. Below: how a model is built, the types compared, what a production model needs, what I modeled for client work, and what to send if you want one made.
Written by Marco Mori, 3D artist, Lindau, Germany · Updated 2026-11-07
What is 3D modeling?
3D modeling is the process of creating a digital object in three dimensions. Wikipedia defines it as developing a mathematical, coordinate-based representation of a surface in a simulated 3D space, by manipulating edges, vertices and polygons. In practice that means an artist builds a shape in software the way a sculptor builds one in clay, except that the result can be turned, lit, animated and rendered.
A model is the first stage of nearly every piece of CGI. Nothing can be animated, lit or rendered until something exists to animate, light and render. That also makes it the stage where mistakes are most expensive, because every later step builds on it.
How is a 3D model made?
The steps differ by object, but the order is stable:
- Collect reference. Photos from several sides, technical drawings or CAD data, real measurements. A model built from memory looks like it.
- Block out the big shapes. Proportion and silhouette first, detail later. If the silhouette is wrong, no amount of detail rescues it.
- Refine the surface. Edges are tightened, curves smoothed, details added where the camera will see them.
- Clean up. Surfaces are tidied so they can be shaded, deformed and rendered without artefacts, and parts are named and sized correctly.
- Hand over for look development. Materials, textures and lighting come next, see what is 3D rendering.
The jump from step two to step three is where beginners and professionals separate. Amateurs add detail early. Professionals lock the shape first, because detail piled on a wrong shape has to be rebuilt.
What are the main types of 3D modeling?
Wikipedia lists polygonal, curve and digital sculpting approaches among the main techniques. In production work they split into four, and one project often uses several.
| Type | How it works | Best for | Weakness |
|---|---|---|---|
| Polygon modeling | Points joined by edges into flat faces, pushed and pulled | Most hard-surface and general work in film, games and ads | Curved surfaces need many faces or smoothing |
| Digital sculpting | Shaping a dense surface like clay with brushes | Skin, faces, creatures, organic folds | Output is heavy and needs cleanup before animation |
| Curve and CAD modeling | Surfaces defined by exact curves, as in NURBS | Engineered products, exact dimensions | Heavy and too plain for close renders without work |
| Procedural modeling | Shapes generated by rules and parameters | Variations, repeating structures, changing specs | Needs setup time, pays off when something changes |
I work across the first three and lean on the fourth because my background is Houdini. When a product spec changes halfway through a job, a procedural build changes with a parameter instead of starting over. The about page explains why that matters.
What does a model need to be usable in a commercial?
A good-looking shape is not the same as a usable model. The question that decides how much work a model needs is simple: how close does the camera get? That is the edge most generic explainers miss.
| Camera distance | What the model needs | Where the effort goes |
|---|---|---|
| Far, small in frame | Right shape and proportion | Almost nothing beyond the silhouette |
| Medium, product hero shot | Accurate edges, correct bevels, clean surfaces | Edge highlights catch light, so bevels must be right |
| Close-up | Fine surface detail, real-world imperfection | Seams, screws, texture, small wear |
| Extreme close-up, skin or hands | Sculpted anatomy and pore-level detail | Most of the budget, and the hardest part to fake |
Two further requirements apply whatever the distance. The model must be the right size, because physics, depth of field and light behave wrongly at the wrong scale. And if it will move, it must be built so it can bend where the real object bends, which is why animation-ready models are planned differently from models made only for stills. The computer animation guide shows where modeling sits in that chain.
What is the difference between 3D modeling and 3D rendering?
Modeling builds the object. Rendering produces the picture of it. Between them sit materials, textures, lights and the camera, which are the steps that make a model look real. A model with no material is a grey clay shape, which is exactly how artists often review it first.
The practical effect is that changes are cheap before rendering and expensive after it. A different shape after the render has started means a new model and a new render. That is why approvals in a 3D product visualization job run in order: shape first, look second, animation third.
What is 3D modeling used for?
Wherever a physical or imagined object has to appear on a screen or be produced. The main fields are advertising and product visualization, film and TV VFX, games, architecture, industrial design and 3D printing. In advertising the model is usually the stand-in for a product that is not finished yet, is too expensive to ship to a shoot, or has to be shown in impossible situations. The comparison to a real shoot is in CGI vs live action product commercials.
What have I modeled for client work?
These facts come from my project pages, not from a template.
| Project | What was modeled | Technique and note |
|---|---|---|
| Trainline Tap That App (2025, via Wieden+Kennedy Amsterdam) | The hyperreal hands | Modeling and sculpting, then look development for skin, nails and the app button. Hands in close-up are the hardest case in the table above |
| PUMA Be the Spark (2020) | A portrait head, a hollow interior and footballs | Modeling, layout, dynamics and timing of the shot. Final compositing was done by Nuderobot |
| Hyperreal portrait studies (2021 to 2025) | A personal series of photoreal heads | ZBrush for sculpting, Houdini for grooming and scene assembly. A test bed for anatomy and surface detail. How it works is covered in how hyperrealistic CGI portraits are made |
Where a client has CAD data, I start from that. Otherwise I model from photos, drawings and dimensions, which takes longer and needs its own approval round before anything is lit.
Which software is used for 3D modeling?
Blender, Maya, Cinema 4D and Houdini cover general polygon and procedural work. ZBrush is the standard for sculpting. CAD packages such as SolidWorks or Fusion create the engineered models that product companies already own. The choice matters less than the handover: a model should leave the modeling tool as clean, correctly sized geometry that the next person can use. My own set-up is Houdini and ZBrush for building, Redshift for rendering and Nuke for the finish. The Houdini side is described in the Houdini FX service.
What should you send if you want a model made?
Four things make the first quote accurate:
- CAD or product data, if it exists. STEP, OBJ or FBX all work.
- Reference photos from every side, plus any measurements.
- The closest camera distance you expect, or an example video with a similar framing.
- Whether the object moves or deforms, such as a hinge, a hand or fabric.
Without those, a model gets priced from guesses. With them, the quote is a short conversation. Write to me with a few lines and you will get a straight answer, including when a ready-made model or a cheaper route would do. The full production scope is described under 3D animation services.
Written by Marco Mori, 3D artist and creative director in Lindau at Lake Constance, Germany, working with brands and agencies worldwide.
Frequently asked questions
What is 3D modeling in simple terms?
It is building a digital sculpture or object in a computer. The result is a 3D model, a shape with width, height and depth that software can show from any angle, animate or turn into a picture.
What are the main types of 3D modeling?
Polygon modeling, digital sculpting, curve based and CAD modeling, and procedural modeling. Polygon modeling is the most common in film and advertising, sculpting is used for organic detail like skin, CAD is used for engineered products, and procedural modeling builds shapes from rules.
What is the difference between 3D modeling and 3D animation?
Modeling creates the object. Animation makes it move. A model can exist and be rendered as a still without any animation, but an animation always needs a model first.
What is the difference between 3D modeling and 3D rendering?
Modeling builds the shape. Rendering calculates the final picture of it, with materials, light and camera. See what is 3D rendering for how that second step works.
Is 3D modeling the same as 3D printing?
No. 3D modeling is the digital design step. 3D printing is a way to make a physical object from a model. A model made for a CGI ad usually needs changes before it could be printed, and a printable model is often too plain for a close-up render.
Do I need a 3D model if I already have CAD data?
Often you have most of it. CAD data is accurate but usually too heavy and too plain for rendering, so it is cleaned up, simplified and given surface detail. That is faster than modeling from scratch.
What software is used for 3D modeling?
Blender, Maya, Cinema 4D and Houdini for general modeling, ZBrush for sculpting, and CAD tools such as SolidWorks or Fusion for engineered parts. I model in Houdini and sculpt in ZBrush.
How long does it take to model something?
It depends on how close the camera gets and how complex the object is. A product seen from far away is quick. A hand or face in extreme close-up is where most of the time goes. A fixed number would be wrong in both directions, so a model gets a time estimate only after a look at the brief and the references.