01
The Linen Veil
A concealment shot. Half the face is hidden, so the half you can see carries everything, and the model stops trying to average out a whole face.
Read the prompt
08Free · seven production prompts · updated 2026
Seven prompts
Seven production-tested image prompts, each with the frame it produced and the reasoning behind every line. Free, no sign-up, and yours to use commercially. The fastest way to explain what a well-built prompt is worth is to hand you seven and let the results do the arguing.
02
The Windswept Field
Negative space as a status signal. The subject sits in the left third and the rest of the frame is weather. Reads as campaign, not as snapshot.
Read the prompt
03
The Paddock
The most structurally demanding prompt in the set: a second living subject, a full seated body from boots to head, and a strict black-and-white exposure.
Read the prompt
04
The Garden, 85mm
The commercial workhorse. Warm, approachable, unmistakably a photograph, and written with variable fields so one prompt becomes a hundred.
Read the prompt
05
The Webcam Selfie
Deliberately imperfect. A slight frame tilt, front-left screen light and a face too close to the lens, the grammar of a real photo taken by the person in it.
Read the prompt
06
The Light Bar
One narrow beam across the eyes and nothing else. Underexpose everything, expose for the strip, and let the frame do the rest.
Read the prompt
07
The Product Hero
Proof the method travels. Same discipline, no human subject: an entire software landing page rendered as a single image.
Read the prompt
Questions
Read them structurally, not literally.
Did you write all of these yourself?
No, and I would rather say so. These are handpicked. Some are mine, some I found and rebuilt until they produced the same result twice, and all of them I reworked to run on a placeholder instead of one fixed subject. Every image on this page came out of the prompt beside it, generated by me, using my own character.
Will they stop working when the models change?
Individual lines will. The architecture will not. Identity locks, labelled blocks and negative constraints have survived every model generation so far. When a specific prompt degrades, the fix is to revise the block that broke rather than throw the whole thing out.
Can I use the outputs commercially?
Your outputs, yes. Client work, your own brand, paid campaigns, resale of the images you generate. Check your generator's licence terms as well. The only thing I ask is that you do not repackage the prompt text and sell it on as your own product. My images are a separate matter: the frames on this page are my own character and they stay mine. Take the prompt, run your own subject.
What do I need to run them?
All seven frames here were made in GPT. The prompts are not tied to it. They carry plain placeholders for the subject, so they run just as well in Midjourney, Nano Banana, Seedream, Flux Kontext or Stable Diffusion XL. None of them require a reference image. Attaching one simply gives the model a face to hold on to, which is what keeps the same person turning up frame after frame.
I write my own prompts.
Then you will get more out of these than most people, because you will read them structurally rather than copy them. The value was never that you cannot write one. It is that you should not have to write the same one for the fourth time.
