Turning an AI haircut render into something your barber can use

You have the render. You show it to the barber, he nods, and forty minutes later you have a haircut that is related to the picture without being it. Nothing went wrong in the shop. The photograph never contained the instruction.

What a photograph does not say

A picture shows a result. A barber works from a method, and the method has four or five decisions in it that are invisible in any photograph. Hand the same image to five barbers and you will get five haircuts, because each one resolves those decisions differently and none of them were asked.

This is not a shortcoming of AI renders specifically. It has been true of every photograph anyone has ever carried into a barbershop. It just matters more now, because the render raises your expectation of how close the result will be.

Guard numbers

This is the unit the trade actually works in. Each number is that many eighths of an inch of hair left behind, which is why a two-number instruction like “#2 blended to #1” carries more information than a paragraph of description.

The four things to specify

• ·How high the fade goes. Low, mid or high, described against a landmark on your own head: at the temple, level with the top of the ear, or above it. This is the decision that changes the haircut most and the one photographs hide worst.
• ·How short the shortest point is. A guard number, or skin. “Faded” without a number is not an instruction, it is a genre.
• ·The length on top, in inches. Not “short”. Short means two very different things to a man with a buzz cut and a man growing out a middle part.
• ·The neckline. Natural, tapered or blocked. Natural follows your own hairline and grows out invisibly. Blocked is a straight horizontal line, looks sharp for two weeks, and then looks like a shelf.

The prompt that writes the spec

Give it the render, or any photograph of a haircut you like, and it converts the picture into the rows a barber needs. The UNKNOWN rule is the important part: without it the model invents a fade height for the back of a head the photograph never showed.

A photograph of a haircut is attached. Write the instruction a barber would
need in order to reproduce it, and nothing else.

Use exactly this format:

SIDES — clipper guard at the top of the fade, guard at the shortest point, and
  where the fade line sits relative to the ear and the temple.
TOP — length in inches, the cutting technique (point cut, scissor over comb,
  clipper over comb), and the direction it is styled.
FRINGE — length relative to the eyebrows, and blunt, textured, or swept.
NECKLINE — natural, tapered, or blocked.
PART — none, natural, or hard.
UPKEEP — how many weeks before it is visibly out of shape.

RULES
If the photograph does not show enough to answer a row, write UNKNOWN for that
  row. Do not guess and do not infer from the front what is only visible from
  the back.
Do not describe the man. Do not comment on whether it suits him.
Do not add advice, alternatives, or a closing paragraph.

What to actually say in the chair

Say the four decisions out loud before you show anything. The photograph is a reference for the shape, not the brief, and a barber who has heard the numbers first will read the picture the way you meant it.

• ·Lead with the fade height. “Low taper, so the fade sits at my temple and stays under the corner.” It is the one that cannot be walked back.
• ·Then the shortest number. “Number two at the shortest, blended up.”
• ·Then the top, in inches. “Leave two inches on top, textured.”
• ·Then the neckline. “Natural at the back, not squared off.”
• ·Then show the photograph. And say what you are showing it for. “This is the shape I am after” is different from “make me look like this”, and barbers hear the difference.

Every cut in the catalog already carries those four rows written out, so you can read them off rather than reconstructing them from a render. That card is the reason this app exists: the render tells you whether you want the cut, and the card is what gets you it.