LarpGPT · Guides
The anatomy of a photo-realistic image prompt: subject, scene, light, lens
How the prompts in LarpGPT are built, what each part does in an image generator, and how to adapt a prompt without breaking it.
One sentence per decision
A photo prompt works best when it reads like a photographer’s brief: one subject, one place, one moment of light, one lens. Vague words (“beautiful”, “amazing”) add nothing the generator can act on; concrete ones (“wet cobblestones”, “late afternoon backlight”) do. LarpGPT’s prompts follow that structure so that each part can be changed on its own.
The subject
Say who or what is in the frame and what they are doing, in a few words. LarpGPT writes subjects as silhouettes or figures seen from behind, which keeps the image about the scene rather than about a specific face. If you replace the subject, keep the same level of detail: “a cyclist in a yellow rain jacket” rather than “a person”.
The scene and the moment
Location, weather and time of day set most of the mood. “A mountain road above the clouds at dawn” tells the generator about colour, depth and haze in one line. Change one element at a time, generate, then compare; changing three at once makes it impossible to know which word did what.
Light and lens
Light words are the most powerful part of a photo prompt: “soft window light”, “hard midday sun”, “neon reflections after rain”. Lens words control the look: a wide angle exaggerates space, an 85 mm portrait lens compresses it and blurs the background, a slow shutter adds motion blur. Keep one light and one lens per prompt.
Copy, adapt, disclose
LarpGPT copies a full prompt to the clipboard in one tap; paste it into the generator you already use, whose terms and pricing apply. Results vary between generators and versions. Where the platform you publish on asks for synthetic images to be labelled, label them, and do not use prompts to imitate a real, identifiable person.