Create
Director-style portrait or product still — lighting diagram in words, lens choice, backdrop, and negative prompts tuned for catalog-clean output.
Directors get better frames.
Pick the shot type, describe the subject in plain photographic language without naming celebrities, choose a lighting recipe and backdrop like you are renting a studio for the hour. Three-point soft, Rembrandt key, high-key bright, low-key dramatic, natural window — each maps to a believable studio setup the model can render. Negative prompts kill plastic-skin beauty filter, watermarks, and extra limbs that haunt stock-gen output. The system prompt enforces respectful representation and refuses celebrity likenesses.
Six photography decisions that change everything.
Each recipe maps to a real studio setup with believable physics.
Default flattering
Key, fill, and rim lights with diffusion — universal portrait setup that flatters most subjects.
Dramatic single key
Triangle of light on the cheek opposite the key source — classic for serious editorial portraits.
Bright catalog
Even, bright, minimal shadow — the look used for clean ecommerce and clinical product shots.
Cinematic shadow
Deep shadows with selective highlight — moody, theatrical, often used for fashion and music.
Editorial warmth
Soft side light from a single source — the look magazines pretend they used "available light" for.
Pre-visualization moments where booking is too expensive to guess.
The difference between a stock photo and a campaign image is vocabulary.
Generic AI image tools accept "professional headshot" as a prompt and produce something that looks generated. Real photographers brief shoots with shot type, lens, lighting setup, backdrop, and wardrobe — every word adds constraint that the model can render. This template forces those decisions up front. The result reads like a photo a hired professional would deliver, not the AI default of plastic-skin perfection. Negative prompts catch the rest: no watermarks, no extra fingers, no uncanny smoothness.
Habits that compound across content production work.
Describe generically — do not upload a stranger's face to clone without consent. The system prompt refuses celebrity likenesses and trademarked characters explicitly.
In most cases yes for original concepts, but rights depend on the underlying image model's terms. Always review the license before campaign use, and avoid the model when the output would be confused with real talent endorsement.
With proper negative prompts ("beauty filter plastic skin, smooth airbrushed face"), modern models render plausible skin detail. Default behavior tends toward over-smoothing.
Describe your style in lighting, lens, and backdrop terms — the more specific your reference, the closer the match. Pair with sample style cues in the subject description.
Specify ethnicity, age, and gender presentation respectfully in the subject description. The model honors specificity but can default to homogeneous outputs without it — be explicit.
Image-generation models tuned for photorealism — defaults vary by capability. Try multiple models for the same brief; quality varies meaningfully across providers for studio work.
Lock everything except the variable you are testing — pose for fashion, prop styling for product, expression for portraits. Re-run with one change per iteration.
Lights cost money — pixels are cheap.
Test moods, lighting recipes, and casting directions before you book photographers and stylists. The pre-visualization saves real-shoot time, gives your creative director something concrete to align around, and prevents the expensive-mistake-on-set conversation.