Polygon and box modelling
Polygon modelling builds a surface directly from vertices, edges and faces. Box modelling is its most common starting technique: begin with a simple primitive such as a cube or cylinder, then extrude, cut and move its parts towards the final shape.
Polygon meshes are the format every renderer and real-time engine works with, which is why most other methods end with a conversion to polygons. Subdivision surfaces let a fairly light quad mesh render as a smooth, detailed form.
Digital sculpting
Sculpting treats the mesh like clay: brushes push, pull, smooth and pinch millions of small faces to build organic forms, soft folds and surface detail. Dynamic topology adds detail only where the brush needs it.
Sculpts are usually far too dense for direct use, so the detail is captured in a normal map and the model is retopologised into a lighter, cleaner mesh for rendering or animation.
NURBS, curves and parametric CAD
NURBS (non-uniform rational B-splines) describe surfaces mathematically rather than as flat faces, so a curve stays perfectly smooth at any zoom. Parametric CAD builds parts from dimensions, features and constraints, so a design can be changed by editing a number.
These methods are the standard for engineering and manufacture because they are precise. For rendering, the surfaces are tessellated into a polygon mesh, and the quality of that conversion decides whether curves facet or stay smooth.
Boolean and procedural modelling
Boolean modelling combines shapes by adding, subtracting or intersecting one volume with another: a fast way to cut holes, slots and recesses in hard-surface objects. The resulting geometry often needs clean-up before it subdivides well.
Procedural modelling generates geometry from rules and parameters, for example with node-based systems. It suits repeated or variable forms, where changing one value regenerates the whole model.
Photogrammetry and 3D scanning
Photogrammetry reconstructs an object from many overlapping photographs; 3D scanners capture its surface directly. Both suit objects that already exist, such as a heritage product or a physical prototype.
Scans produce dense, uneven meshes that usually need retopology, and reflective or transparent materials scan poorly. A scan records the object as it is, which is not always the finish a product launch needs.
Which types product CGI uses
Product CGI usually begins with the manufacturer’s CAD, because it carries the exact dimensions. The CAD is converted and cleaned into a polygon mesh, with detail added where close-ups need it, then textured, lit and rendered.
Sculpting appears where a product has organic or soft elements, and scanning where only a physical object exists. A general-purpose package such as Blender or 3ds Max covers polygon modelling, sculpting and procedural work in one place.