Sarajevo · working worldwide .shock Suite — software for photographers →

Political portraits — 07

The machine does the counting. A person signs the delivery.

Five crops per person across four hundred people is two thousand files that all have to be cut identically. That is a machine’s job. Deciding whether the result is good enough to put on a billboard is not — so the pipeline is automated end to end and gated by a person before anything leaves the studio.

Included in every tier 5 crops computed per person 100% human-reviewed

What this service delivers

This page describes the delivery pipeline rather than a separate product, because it is not sold separately — it is how every tier on this service is delivered. Frames come off the card and are grouped by person automatically. A face detector returns the face box and landmarks for each selected master, and every output crop is computed from the eye line outward at a face-to-frame ratio set per format, so the ballot frame, the avatar and the six-metre billboard all place the head in the same relationship to the edges. Naming follows your convention rather than ours. Colour and density are checked as values against the spec-day reference rather than by eye. Then a person opens the set, compares it to the call sheet, and either signs it or sends it back. The automation buys speed and consistency; it does not buy the right to skip that step.

Crops computed

5 per person

Sorting

Identity-matched

Human review

100%

Cost

Included

What the pipeline does

Automated where it is arithmetic. Manual where it is judgement.

01

Face-anchored crop geometry

The detector returns a face box and landmarks; each crop is computed from the eye line at a face-to-frame ratio set per output shape, so every format frames the head consistently.

02

Shape-aware headroom

A short 16:9 or 3:1 cut from a tall portrait runs out of room above the crown first. The geometry grows the crop rather than clipping the head to hit the ratio.

03

Identity-matched sorting

Faces are embedded and frames grouped by person, so a multi-thousand-frame slate becomes per-person folders without anybody scrubbing through thumbnails.

04

Naming to your convention

Files are named against your list and your structure at export — not renamed by your comms team the week of the deadline.

05

Numerical colour QA

Background density and key exposure are compared as values against the spec-day reference, which catches the third-of-a-stop drift that no one sees in a room.

06

A delivery manifest

A per-person checklist of every expected file, so a missing crop is caught by a list rather than by whoever needed it first.

What the automation is never allowed to do.

The line

Every capability on this page is about moving files accurately at volume. None of it touches what the person in the frame looks like — and that boundary is written into the contract rather than left to good intentions.

  • No faces are restructured

    No reshaping, no body alteration, no change to apparent age — by hand or by model. The retouch limit is identical whether a person or a tool does the work.

  • No generated people, no generated backgrounds

    Nothing on a delivered political portrait is synthesised. The background is a background that was in the room, and the person is the person who sat in front of it.

  • Nothing ships unreviewed

    Automation drafts; a person approves. If the review pass cannot be done in the time available, the deadline moves — the review does not get skipped.

How it runs.

Step 01

Ingest & group

Frames come off the card, are backed up twice, and are grouped by person automatically before anybody starts looking at them.

Step 02

Select & finish

A human selection per person, retouched inside the written limit, colour-matched across the whole set against the spec-day reference.

Step 03

Compute & check

Every crop computed from the approved master, named to your convention, checked against the delivery manifest — then reviewed and signed off by a person.

What you receive

Automated

  • Face-anchored crops in all five formats
  • Identity-matched per-person folders
  • Naming applied to your convention
  • Numerical colour and density QA

Human

  • Selection of the master frame per person
  • Retouch inside the written contractual limit
  • Review of every generated crop before delivery
  • Set checked against the call sheet

Governance

  • No face restructuring, by hand or by model
  • No synthetic people or generated backgrounds
  • Separate encrypted storage per engagement
  • Agreed purge date, confirmed in writing
Disclosure

Every frame on this page is our own studio reference work, photographed to demonstrate the standard. No client campaign material appears here, or anywhere else on this site.

Before you brief.

Is any part of the delivered image generated by AI?

No. The automation locates faces, computes crop geometry, groups frames by person, applies naming and checks colour numerically. It does not generate, alter or invent any part of the picture, and no delivered political portrait contains synthetic imagery.

Why automate the crops at all?

Because two thousand hand-cut crops are not consistent, and the inconsistency is invisible until the whole slate is laid out together. A computed crop places the eye line in the same position on every person in every format, which is exactly the thing a human doing it four hundred times cannot hold.

What happens when the automation gets it wrong?

A person catches it, because every generated file is reviewed before delivery. Difficult frames — unusual poses, headwear, more than one face in shot — get re-cut by hand. That is a normal part of the job, not an exception.

Included in every tier — not a line on the invoice.

The pipeline is how the work is delivered, not an upgrade. Express 24-hour delivery on a whole batch is available at +40%.

See the full rate card

Ask us how the files will arrive.

Naming convention, formats, folder structure, deadline. Tell us what your systems need and the delivery is built to match it.