One photo, 6 camera angles: 6 ChatGPT prompts to re-shoot it

Re-shoot the same photo as a drone shot, a fisheye or a telephoto candid, and it looks like a professional photo shoot. This is the full version of the 6 ChatGPT prompts from the Reel, with why they are written this way and how to stop the person from changing.

01Get ready: one good original photo

A vertical photo with the person in the middle. Full body or half body both work.
Natural light and a clear face. No sunglasses, and do not crop the face.
Keep the background simple. It helps the AI keep the details.
Open ChatGPT (the free plan can also make images, with a limit) and upload the photo.

Sample original: a woman with long hair sits on a grassy slope under a big tree by a lake, shot at eye level — The original. This is an AI-generated person, not a real one. All 6 photos below come from this one.

02The base sentence

Start with the same English sentence every time and only change the ___ part. The fixed part tells the AI not to change the person, face, outfit, pose or place.

Base sentence

Re-shoot this photo as ___. Keep the same person, face, outfit, pose and place. Photorealistic, 4:5, no text.

036 angles to fill in the ___

1 Drone

a top-down drone shot from 15m above

Drone version: shot from straight above, the person is small and the tree and shore become shapes — 1 Drone: the person gets small, and the tree, grass and water edge turn into shapes.

2 Fisheye

a 10mm fisheye shot close to the face, circular frame

Fisheye version: a circular frame close to the face, with the tree and hills curving around the edge — 2 Fisheye: a circular frame with the face in the middle, and the tree and hills curve around it.

3 Telephoto candid

a candid 200mm telephoto shot through blurred grass

Telephoto version: blurred tall grass in front, the person far away on the slope — 3 Telephoto: blurred grass in the foreground, like a stranger took it from far away.

4 Film

35mm Kodak Portra 400 film, grain, light leak

Film version: warm tones, visible grain and a light leak on the left edge — 4 Film: warmer colours, grain, and a light leak on one edge.

5 Golden hour backlight

golden hour, sun behind the subject, rim light in the hair

Golden hour version: the sun behind her, a glowing edge on her hair, the grass lit up — 5 Golden hour: the sun is behind her, her hair has a bright rim and the grass glows.

6 Night flash

a night shot with direct on-camera flash

Night flash version: dark background, the person lit by a direct flash from the front — 6 Night flash: the background goes dark and a direct flash lights her up, like a party snapshot.

For example, the full prompt for number 3 is: Re-shoot this photo as a candid 200mm telephoto shot through blurred grass. Keep the same person, face, outfit, pose and place. Photorealistic, 4:5, no text.

04Why it is written this way

Photographer language (focal length, camera, light position) works better than adjectives. "200mm" is far more precise than "shot from far away".
"Keep the same person, face, outfit, pose and place" is the most important part. Without it, the person changes.
"4:5" is the Instagram post ratio. "no text" stops the AI from adding its own words.
Each prompt changes only one thing: the angle, the lens or the light. Change too much at once and the original features drift.

05Watch out

If you don't like the result, just say: "The face doesn't match. Try again, the face must be the same as the original."
Fingers, text and logos are where AI makes the most mistakes. Zoom in and check.
Read OpenAI's terms of use before any commercial use. It works well for pitch mock-ups and social post references.