Material, Lighting and Realism in CGI Product Video

Why realism in CGI comes from surfaces and light rather than geometry, and the specific decisions that make a rendered product look manufactured.

Clients reviewing early CGI usually comment on the model, and the model is almost never the problem. Geometry is the easy part of product visualisation: a STEP file from engineering is already dimensionally correct. What separates a render that looks like the product from one that looks like a video game asset is the behaviour of the surfaces and the quality of the light, and both are craft decisions rather than technical ones.

Material realism comes from getting three things right: how much light a surface reflects, how sharply it reflects, and how the reflection changes with viewing angle. A powder coated housing reflects a little, diffusely, with a soft falloff. Brushed aluminium reflects strongly in one direction and weakly across it, producing the characteristic stretched highlight. Injection moulded plastic scatters light slightly beneath the surface near thin edges. Getting these behaviours right matters more than the colour, which is what clients usually focus on.

Imperfection is the single most effective realism technique and the one most often omitted because it feels like adding a defect. Real manufactured objects have microscopic surface variation, slight orange peel in the paint, fingerprints, dust, faint scratches from handling, edge wear where parts meet. A perfectly uniform surface reads as computer generated because nothing physical is uniform. Adding controlled, subtle imperfection is what makes an object look made rather than modelled.

Edges carry disproportionate information. Real objects have no perfectly sharp edges; every edge has a small radius from tooling or finishing, and that radius catches light and produces a bright line that defines the form. CGI with mathematically sharp edges disappears at those transitions and looks flat. Adding a small bevel everywhere is a modelling step that costs little and improves realism more than almost any material adjustment.

Lighting should be built from a real setup rather than invented. The arrangement that flatters manufactured objects is consistent: a large soft key source establishing form, a controlled fill preventing the shadow side from going black, and a rim or edge light separating the object from the background and defining its silhouette. Most memorable product imagery is a variation on this. Flat, even lighting, which is the default in an unconsidered render, makes an object look like a catalogue entry.

The environment does much of the work for reflective products because what appears on a glossy surface is mostly the surroundings. A product rendered against an empty grey void has nothing to reflect and looks dead. Even for a shot that reads as a plain background, the object needs a constructed environment around it to produce plausible reflections. This is why studios in this field invest in environment libraries rather than in more detailed models.

Colour behaviour deserves formal verification rather than eyeballing, because brand colours are the detail clients check. A specified print or paint reference does not map directly to a rendered surface under a particular light, and the same value reads differently against different backgrounds. Jonauskaite et al. (2020) documented consistent patterns of emotion associations with colours, which is a reason to treat the palette as a deliberate decision, and the practical step is to render a calibration frame, compare against a physical sample, and lock the material.

Depth of field and lens character should be chosen rather than defaulted. Real product photography is shot on real lenses with real optical behaviour: a slight falloff at the frame edges, chromatic aberration at high contrast boundaries, a specific bokeh shape from the aperture blades. Rendering with a mathematically perfect virtual lens produces an image that is subtly unlike any photograph the viewer has seen, and adding modest optical imperfection closes the gap.

The commercial argument for this level of care is that the render is doing the job a photograph would do. Poushneh (2021) found that perceived proximity to a virtual product influenced purchase intention, and proximity depends on the object reading as physically present. Johnson Jorgensen and Sorensen (2026) similarly documented that dimensional presentation shapes how products are perceived. A render that reads as a diagram creates understanding; one that reads as a photograph creates desire.

The practical review sequence that catches most problems is to check the grey model for form and proportion first, then the materials against physical reference photographs under matched lighting, then the lighting for shape and separation, then the lens and grade. Reviewing everything at once, on the first full render, produces feedback that mixes all four and is difficult to act on.

References

Jonauskaite, D., Parraga, C. A., Quiblier, M., & Mohr, C. (2020). Feeling blue or seeing red? Similar patterns of emotion associations with colour patches and colour terms. i-Perception, 11(1), Article 2041669520902484. https://doi.org/10.1177/2041669520902484

Poushneh, A. (2021). How close do we feel to virtual product to make a purchase decision? Impact of perceived proximity to virtual product and temporal purchase intention. Journal of Retailing and Consumer Services, 63, Article 102717. https://doi.org/10.1016/j.jretconser.2021.102717

Johnson Jorgensen, J., & Sorensen, K. (2026). Millennial perceptions of augmented reality in retail. Virtual Worlds, 5(3), Article 30. https://doi.org/10.3390/virtualworlds5030030