A synthetic image that survives a second look
The first look is easy now. The studio's whole discipline lives in what happens on the second.
Generated images pass the first look almost universally. Scroll speed is not a test any more. The test is the second look — the moment a viewer stops, and their eye goes looking for the thing that is wrong.
It usually finds one of four things: a specular highlight that is too clean for the material it sits on, a shadow that does not agree with the key light, a surface with no history, or a depth of field that behaves like a filter rather than a lens. None of these are model failures exactly. They are the absence of a decision.
The discard rate is the work
In the Lab, a material study routinely produces two hundred frames and keeps four. We publish the discarded frames next to the survivors with a written verdict on each, because the verdict is the transferable part. Anyone can generate two hundred frames. Knowing which four are load-bearing is the skill a client is actually buying.
If a frame reads as an effect, it is thrown away. That single rule removes most of the output.
Three rules we keep applying
- Commit to one light. Two believable sources are harder to hold than one, and most synthetic images fail because nobody decided where the light was coming from.
- Give the surface a history. Wear, dust, a fingerprint, an edge that has been handled. Perfect materials read as renders because perfect materials do not exist.
- Let the lens behave. Real optics fall off, flare imperfectly, and lose the corners. A frame with no lens in it looks like a picture of nothing in particular.
Why we keep doing this without a client
The Lab is not a portfolio exercise. It is where the failure modes are found cheaply, so they are not found expensively during a commercial. The lighting reference we use on interiors came out of a study nobody commissioned. So did the grade sitting under most of our moving image work.
Every piece of self-initiated work eventually pays for itself in a client engagement — usually as a decision made in an afternoon that would otherwise have taken a week of the client's budget to discover.