How to Write a Maternity Photoshoot Prompt for ChatGPT That Looks Professional

The short version: ChatGPT can produce a convincing maternity photoshoot image if you feed it lighting, lens, pose, wardrobe and setting details the way you'd brief a real photographer, not a vague wish. It still struggles with hands, fabric folds and matching faces across a set, so treat it as a mood board and marketing tool rather than a replacement for an actual shoot. I tested this with a client's business and got a 15% usable rate out of 40 images, which tells you exactly how much trial and error to expect.

Why most people get this wrong from the first sentence

I had a client last spring who runs a maternity and newborn photography studio in Surrey. Bookings were thin in February, her slowest month every year, and she wanted to fill her Instagram grid with a few "styled" maternity images to keep the feed alive between real client shoots. Her first prompt was something like "pregnant woman, golden hour, elegant, professional photo." She got back images that looked like stock photography from 2016. Flat, generic, a bit uncanny around the hands, and nothing that matched the soft, editorial style her actual client photos had.

That's the mistake almost everyone makes with ChatGPT image prompts, not just for maternity shoots. You describe the subject and the mood but not the camera. A real photographer never just says "make it look nice." They think in focal length, aperture, light source and direction before they ever press the shutter. ChatGPT responds to the same language because the model was trained on millions of images that came with exactly that kind of caption data.

The prompt structure that produces usable images

After a few rounds of testing with that client, we settled on a five-part structure. Every prompt we wrote followed this order, and it cut our wasted generations by more than half.

• Subject and pose: who, roughly what body position, what she's doing with her hands (holding the bump, resting a hand on a wall, arms crossed loosely)
• Wardrobe and fabric: specific garment type and material, not just "dress"
• Setting: indoor studio, outdoor field, home nursery, with one or two physical details
• Light: named light quality (window light, golden hour, soft studio strobe) and direction
• Camera language: lens length, aperture, film-style reference if you want a specific look

Here's an actual prompt from that batch that worked:

"A pregnant woman in her third trimester standing side-on by a large window, wearing a fitted ribbed jersey dress in sage green, one hand resting gently under her bump, soft diffused natural window light from the left, shallow depth of field, shot on an 85mm lens at f/1.8, warm neutral tones, editorial maternity photography style, no text, no watermark."

Compare that to "pregnant woman, elegant, golden hour." Same intent, wildly different output quality. The second one gives ChatGPT nothing to anchor to, so it defaults to the most generic version of "pregnant woman photo" it has ever seen.

The specific details that separate amateur from professional-looking

Once the basic structure is in place, the difference between a passable image and one that looks like it came from a paid shoot sits in five small details.

• Name the fabric, not just the garment. "Silk slip dress" or "ribbed cotton" reads completely differently to the model than "dress." Fabric words trigger texture rendering that flat clothing terms don't.
• Give the hand something to do. Hands are still the weakest part of any AI image generator, including the one inside ChatGPT. If you don't specify a pose, you'll get strange, extra, or fused fingers more often. Telling it exactly where the hand goes and what it's touching reduces this, though it doesn't eliminate it.
• Pick one dominant light source and describe its direction. "Soft light" alone is too vague. "Late afternoon sun coming through a sheer curtain from the right" gives it something concrete to render shadows against.
• Use real photography style references. "In the style of natural light maternity portraiture" or "editorial fashion photography, muted colour grade" works better than adjectives like "beautiful" or "stunning," which the model has seen attached to millions of unrelated images.
• Say what you don't want. Adding "no text, no watermark, no distorted hands, natural proportions" at the end of the prompt reduces the error rate. It's not foolproof, but it's a free line that costs nothing to add.

What ChatGPT still can't do, and why the marketing around this tends to oversell it

Here's the part most guides on this topic skip past because it's less exciting: ChatGPT's image generation is good at a single, striking image. It is not good at a consistent set. If you need six images of "the same woman" in different poses for a portfolio page or a campaign, you will not get one consistent face across all six. The model doesn't hold identity the way a camera does. My client's studio ran into this directly. We tried to build a mini "styled shoot" of four images meant to look like one session and every single face came back slightly different. Different nose, different jawline, different skin tone in two of them. We ended up using only the single best image from each generation batch rather than trying to build a matching set.

Out of 40 total generations across that project, only 6 were usable without further editing. That's a 15% hit rate. Four more were fixable with a quick crop or a colour tweak in Canva. The rest went in the bin. If anyone tells you they get clean, publication-ready maternity images on the first try every time, they're either exaggerating or working with a much narrower, simpler brief than "photoshoot."

The other thing worth saying plainly: if you're an expectant mother wanting AI-generated images of yourself pregnant, ChatGPT can't do that from a text prompt alone, and even with a reference photo uploaded, the result won't be you, it'll be a plausible-looking stranger who resembles the brief you gave it. That's fine for a mood board, a save-the-date graphic, or a blog header. It is not a substitute for booking an actual maternity photographer if what you want is a real memory of your real body at this stage of your life. The people who get the most value out of this technique are photographers filling content gaps, marketers building campaign visuals, and small studios testing style directions before a real shoot, not families trying to skip paying for one.

A prompt template you can copy and adjust

Once you have the structure down, most of your work is swapping out details for your specific brief. Here's the skeleton we ended up using for that client's studio, which you can adapt for your own setting:

"A pregnant woman in her [first/second/third] trimester, [pose and hand placement], wearing [specific garment and fabric] in [colour], set in [location, one or two physical details], lit by [light source and direction], shot on a [lens length]mm lens at [aperture], [colour tone: warm, cool, muted, high contrast], in the style of [photography style reference], no text, no watermark, natural proportions."

Change the bracketed sections and generate three to four variations per session rather than one. ChatGPT gives noticeably different results even with an identical prompt run twice, so treat each batch like a Polaroid contact sheet: you're picking the best of several, not expecting perfection on the first attempt.

Where this fits into a real business, not just a fun experiment

If you run a photography business, a birth and baby brand, or you're a marketer building a maternity or family-focused campaign, this technique earns its place as a tool for filling gaps between real shoots, testing a colour palette or set design before you book a studio, or building social content when your calendar is quiet. It does not replace a client shoot, and using it to imply you photographed someone you didn't will damage trust faster than any slow month ever will. Label AI-generated maternity images as such if you post them commercially. It's a small line of text and it protects the credibility you've spent years building.

If you're trying to work out where AI tools like this fit into your wider marketing without wasting hours on trial and error, that's the exact kind of thing worth getting a second pair of eyes on. I do this as part of my AI consultant for small business work, mostly because clients waste far more money guessing than they would spend getting the workflow set up right the first time.

For the rest of the series, see the ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps. To write your own, try the free tool that writes the prompt for you.

Related: chatgpt: guidelines and how to pitch.

Frequently asked questions

Can ChatGPT generate a realistic image of a specific pregnant person from a photo?

Not accurately. Even with a reference photo uploaded, ChatGPT's image tool produces someone who resembles the brief, not the actual person, because it isn't designed to replicate a specific individual's face. Use it for mood boards and style references, not as a stand-in for a real photo of yourself.

Why do the hands in my ChatGPT maternity images look wrong?

Hands remain one of the hardest things for AI image generators to render correctly, including the model inside ChatGPT. Giving the hand a specific job in your prompt, such as "resting under the bump" or "holding the edge of a cardigan," reduces errors, and adding "natural hands, no distorted fingers" as a negative instruction at the end of the prompt helps further, though it won't fix it every time.

How many images should I expect to generate before getting one usable result?

Based on real testing across 40 generations for a client's maternity photography business, only about 6, roughly 15%, came out fully usable without editing. Plan for a batch of three or four per prompt variation and expect to pick the best rather than the first.

Is it okay to use AI-generated maternity images in a photography business's marketing?

It can work well for filling content gaps or testing a style direction, but it needs to be labelled clearly as AI-generated if there's any chance a viewer would assume it's a real client photo. Passing off AI images as real client work is a trust risk that isn't worth the short-term content boost.

Useful references

• OpenAI's own platform documentation