Cinematic Photo Prompt Expert
Image prompts that read like real film stills: shot, lens, light, grade and ratio
For filmmakers building mood boards and pitch decks, photographers, designers, writers and anyone who wants AI images that look like a frame from a real film.
Adding "cinematic" to a prompt usually produces a glossy stock photo. This skill makes your agent think like a cinematographer instead: it finds the story moment, then sets five dials deliberately - shot size and angle, lens and aperture feel, light that comes from a source in the scene, a grade described by its traits, and a true cinema aspect ratio - and writes the prompt the way your image model reads it.
It includes starting recipes by genre and mood (neo-noir, classic noir, golden-hour nostalgia, intimate drama, desert epic, gritty realism, horror, deadpan symmetry, dreamy romance, cold thriller), and adapts the result for OpenAI GPT Image 2.5, Google Nano Banana, ByteDance Seedream 5.0, FLUX 3 and FLUX.2, and Midjourney V8, with valid size, aspect-ratio or --ar settings and crop guidance for each cinema ratio.
You get: the skill instructions with a decision table, output format and self-check; a cinematography vocabulary (shot sizes, angles, lens looks, lighting setups, 18 film looks, film-stock characters, texture cues and a ratio table); model adapter notes; and three worked examples.
Looks are described by visual traits, never by copying a specific film, naming a living cinematographer or imitating a real actor, so the images you make stay yours. Model options change often - if a setting is rejected, check your provider's current docs. Results depend on the model you use; review outputs before publishing.
Highlights
- Look recipes built from shot, lens, motivated light, grade and ratio
- 18 film looks described by visual traits, plus film-stock characters
- Cinema ratios (1.37 to 2.76:1) mapped to valid settings per model
- Adapters for GPT Image 2.5, Nano Banana, Seedream 5, FLUX 3, Midjourney
- Original by design: no film copying, living DPs or real likenesses
Example
A cinematic shot of a tired detective in a diner at 3am, rain outside. For a short-film pitch deck, widescreen. GPT Image.
**Look:** Neo-noir neon, medium close-up, 2.39:1 - cold-and-warm light sells the 3am mood. **Model:** GPT Image 2.5 Sunburst **Settings:** size 2752x1152, quality high **Prompt** Use: cinematic film still for a short-film pitch deck, 2.39:1 widescreen. Scene: a detective in her late forties sits alone in a diner booth at 3am, a case file open in front of her, coffee gone cold. She has just read something troubling and stares past the camera at the rain. Camera: medium close-up from slightly below eye level, subject on the right third, rain-streaked window filling the left. 50mm lens perspective, shallow depth of field. Lighting: cold fluorescent overhead, a red-and-teal neon sign glowing through the wet glass and reflecting on the formica, warm amber spill from a pie case behind her. Look: neo-noir colour - inky blacks, saturated neon, cool skin with warm edge light; fine film grain, soft halo around the neon. Constraints: she does not look at the camera; no readable text, logos, watermarks or letterbox bars.
Indie film poster vibe: two brothers sitting on the hood of an old car at sunset in the desert. Midjourney.
**Look:** Golden-hour naturalism, wide low-angle frame, 2:3 poster with title space. **Model:** Midjourney V8.2 cinematic film still, two brothers in their twenties sitting on the hood of a faded turquoise 1970s sedan, the older one pointing at the horizon, the younger laughing mid-sentence, desert highway shoulder at sunset, wide shot from a low angle, 35mm lens perspective, deep focus, low sun behind them with warm lens flare and glowing rim light, honey and dusty-rose sky, long shadows toward the camera, lifted soft blacks, fine film grain, large empty sky in the top third --raw --ar 2:3 --stylize 100 --no text, watermark, letterbox bars **Iterate with:** add --sref from your favourite result to keep the grade across a poster series.
Make it look like [a famous sci-fi film] shot by [a living cinematographer]: a woman walking through a huge empty concrete hall.
I described the look you're after - monumental concrete, haze, a single cold light source, a tiny figure, a restrained palette - instead of naming the film or cinematographer, so the image stays original. **Look:** Large-format epic + clinical cool, extreme wide. **Model:** Nano Banana 2.1 **Settings:** aspect_ratio "21:9", image_size "2K"; crop ~2% of the height for 2.39:1. Create a cinematic film still in a 21:9 widescreen frame for a science-fiction mood board. A lone woman in a long sand-coloured coat walks across a vast, empty brutalist concrete hall, tiny in the lower centre of the frame. Extreme wide shot from eye level, wide lens, deep focus, columns receding in strict symmetry. A single tall slot of pale daylight cuts through thick haze from high on the right...
What's inside
cinematic-photo-prompt-expert/ ├── agents/ │ └── openai.yaml ├── references/ │ ├── cinematic-vocabulary.md │ ├── model-adapters.md │ └── worked-examples.md ├── LICENSE.txt ├── README.md └── SKILL.md
Install by unzipping into your agent's skills folder. Install guide →