How Are Hyperrealistic CGI Portraits Made?

What actually makes a CGI portrait look real?

Not detail density. Almost every CGI head that fails is failing on light behaviour rather than geometry. Skin is a translucent, layered, uneven material, and the eye is astonishingly good at spotting when it is being treated as a smooth plastic surface with a texture map on top.

Three things carry most of the realism:

• Subsurface scattering that varies across the face. Ears and nostrils transmit far more light than a forehead.
• Two specular layers. One broad and soft for skin moisture, one tight for oil and sweat. A single specular layer is the most common cause of the "wax" look.
• Breakup at every scale. Pores, fine wrinkles, uneven colour, small blemishes, asymmetry between the two halves of the face.

The portrait series I have been building since 2021 exists partly as an argument against symmetry and cleanliness. The heads are surreal, but the surfaces have to be believable or the distortion has nothing to push against.

How is a hyperreal 3D head made, step by step?

The ordered process is listed above the FAQ. Two notes about the order that are easy to get wrong:

Skin look development happens before grooming, not after, because hair changes how you read the skin underneath and you will chase your own tail if you groom onto an unfinished shader. And lighting is set before the final texture pass, because a texture that looks perfect under flat lighting can fall apart under a single hard key.

Which software is used for hyperrealistic CGI portraits?

Tool	Used for	Why it is the choice
ZBrush	Sculpting anatomy and fine detail	Handles very high polygon counts for pore-level displacement
Houdini	Grooming, simulation, scene assembly, FX	Procedural, so a change upstream propagates instead of forcing a redo
Redshift	Shading and rendering	GPU rendering with a mature subsurface and hair model, fast enough to iterate
Nuke	Compositing and grade	Pass-level control, lens character, final integration
Substance / texture painting	Skin colour maps and imperfection layers	Layered maps that survive close crops

None of this is prescriptive. A Blender and Cycles version of this pipeline produces the same result with different friction; the comparison of Houdini and Blender covers where each one hurts.

Why do CGI faces fall into the uncanny valley?

Because realism raises expectations faster than it satisfies them. A stylised character is judged as a drawing. A photoreal one is judged as a person, and the viewer starts applying assumptions about eye movement, skin translucency and micro-expression that a still image can dodge but animation cannot.

Practically, the most common triggers are: eyes that are too dry and too symmetric, skin that scatters uniformly, hair that is too even, and motion that lacks the tiny involuntary movement real faces never stop making.

How much does the eye matter?

Disproportionately. Eyes are wet, refractive, never perfectly aligned, and always slightly in motion. A portrait with excellent skin and dull eyes reads as a mannequin; a portrait with mediocre skin and correctly built eyes often passes. It is the highest return per hour in the entire process.

Can you make a hyperreal portrait without a 3D scan?

Yes. Scans give you real asymmetry and correct proportion for free, which is why production pipelines use them, but they also anchor the result to a specific person. Sculpting from photo reference costs more time and buys creative freedom, which is the trade I make in personal work.

What is the fastest way to make a CGI head look worse?

Four reliable methods, all of them common:

• Perfect symmetry. Mirroring the sculpt saves an hour and costs the whole illusion.
• Even lighting. A softbox on each side flattens every structural cue in the face.
• Clean output. No grain, no chromatic aberration, no depth of field. Technically superior, visually synthetic.
• Too much displacement. Overscaled pore detail reads as orange peel, not skin, and it is the classic overcorrection after someone is told their skin looks like plastic.

How does this feed into commercial projects?

Directly. The PUMA Be the Spark film opens on a close-up portrait of Marco Reus before it goes somewhere absurd, and that opening only works because the head holds up under scrutiny. Character-led commercials, campaign key visuals and music video sequences all sit on the same foundation. What starts as personal look development in the case studies ends up as a client capability.

What does this cost to produce?

There is no useful single number, because the drivers are whether a scan exists, how much grooming is needed, and whether the result moves. The general cost logic for CGI work is broken down in what drives the cost of a CGI commercial.

If you have a character-led project in mind, the FX and look development side is described under Houdini FX, or you can describe the idea directly.