ChatGPT or Gemini vs Graydrop for a stacked portrait

You can make a stacked black and white family portrait for free in ChatGPT or Gemini: upload the photo, paste a detailed prompt, and keep regenerating until it is right. The free route is enough when you have time to iterate and can judge every face yourself. Graydrop is worth $12.99 when you want it right without the iterating, with each face checked against your photo automatically.

We make Graydrop, so read this knowing that. It is written to be accurate rather than persuasive, and the free method is published in full on this site: the stacked family portrait prompt is the exact prompt, with the four reasons it fails and how to fix each.

1. Open ChatGPT or Gemini. Both can edit a photo you upload, and both offer image generation on their free plans, with usage limits that each company changes from time to time. 2. Upload the family photo — the clearest one you have, every face visible. 3. Paste the full prompt from the prompt guide. A one-line request ("make this a black and white stacked portrait") almost never works; the detail is what holds the composition together. 4. Check the result against the original, face by face. Regenerate or correct with a follow-up message when something is wrong. 5. Download the version you are happy with.

The same four failures account for nearly every bad result, in our experience of running this prompt at scale:

• —The stacking order comes out wrong. People end up side by side, or in a pyramid, instead of one behind another.
• —The background comes out gray instead of black, which makes the whole portrait look flat.
• —Faces stop being the right people. The model produces a family that looks related to yours rather than your family — the failure that matters most, and the hardest to spot on a small screen.
• —Existing headwear survives when it should have been removed.

None of these is rare. Even with a carefully written prompt, in our production runs roughly one attempt in three needs a retry for some measurable defect. In a chat app, you are the one who has to notice the defect and ask again.

It runs the same kind of image model, inside a process built to catch those four failures before you see anything:

• —Faces are written down first. A vision pass describes each person's face in words, and that description goes into the image prompt.
• —Every result is scored against your original. If the likeness scores below 75 out of 100, it regenerates automatically.
• —The composition is checked, so a side-by-side result or a gray background is caught and redone rather than shown to you.
• —You see the finished portrait, watermarked, before paying. $12.99 unlocks that same image at full resolution, with one free regeneration on request.

It also has limits the chat apps do not: it makes one style of portrait (monochrome and stacked, plus a Christmas version), people only — no pets — and two to six people.

Use ChatGPT or Gemini if most of these are true:

• —You have half an hour and enjoy iterating.
• —The photo has two or three people, all facing the camera in good light.
• —You can study every face closely and would notice if one drifted.
• —It is for a post on social media rather than a print or a present.
• —You want a style Graydrop does not make — color, a different backdrop, your dog.

That is a perfectly good way to get one, and the prompt guide gives you the best starting point we know of.

• —You want it right without the back-and-forth.
• —There are four to six people, which is where order and likeness errors multiply.
• —It is a gift or a print, and a face that is slightly wrong would spoil it.
• —You would rather not judge a child's likeness on a phone screen yourself.

See a black and white family portrait from a photo for how it works, the price, and who it suits and does not.