Diwali Family Photo Prompt for Gemini (Group Portrait)
Family portraits are where AI editors fall apart — three faces in, someone gets swapped or blended. This copy-paste Gemini prompt is built for the multi-face problem: you name each uploaded photo before generating ('photo 1 is the father…'), and the lock keeps every face identical to its own photo — your whole family in festive wear around the puja altar, in one paste. Terrace and three-generation variations are below.
The whole family in one Gemini prompt: parents and kids in festive ethnic wear around the puja altar, every face locked to its own photo — no swaps.
Use the uploaded photos as identity references: photo 1 is the father, photo 2 is the mother, photo 3 is the daughter, photo 4 is the son. Keep every face 100% identical to its own photo — do not change facial features, skin tones, hairstyles, or expressions, and do not swap or blend faces between people. Create an ultra-realistic Diwali family portrait in a living room decorated for Laxmi Puja: the father in a maroon kurta with a Nehru jacket, the mother in a deep green silk saree with a gold border, the daughter in a mustard-yellow lehenga, the son in a cream kurta pajama. They stand close together before a small home altar with marigold garlands, rows of lit clay diyas at two heights, and brass puja plates, warm fairy lights glowing along the wall behind them, everyone smiling gently at the camera. Warm golden cinematic light from the diyas and room lamps, natural skin texture for every person, shallow depth of field, photorealistic, 8K quality. No text, no watermark, do not beautify or slim any face, no extra people.
See what this prompt creates
Every Diwali, the family portrait is the photograph that matters — the one that goes to relatives and gets framed. It's also the one nobody can take — someone is always holding the camera — so families are building it in Gemini instead. The catch is identity: past two faces, image models start swapping features between people. This prompt is written specifically to stop that.
How to use it
- Gather one clear photo per person — front-facing, well-lit, no group shots.
- In Gemini, attach all the photos, then paste the prompt — the opening lines assign each photo to a person, so don't skip or reorder them.
- Generate, then zoom into every face one by one — swaps hide at thumbnail size.
- Fewer or more people? Edit the name list and outfit sentences to match.
Why this prompt works
Two mechanisms do the heavy lifting. The first is naming the photos. Without labels, Nano Banana assigns faces by resemblance; with "photo 1 is the father", it binds each identity to a role and holds it. The second is the lock's second half — "do not swap or blend faces between people". Distinct outfit colors finish the job — maroon, green, mustard, cream. The altar is specified object by object — diyas at two heights, marigold garlands, brass plates.
Two more family frames
- Terrace sparklers: swap the scene for "standing together on the home terrace at night, each adult holding one lit sparkler at arm's length, the children in front holding diya plates, fairy lights along the railing and the city glowing far below".
- Three generations on the sofa: "grandparents seated in the center of a sofa, parents standing behind, children kneeling in front with brass plates of lit diyas, the puja altar visible at the side, warm lamplight". Seating grandparents anchors the composition and gives the oldest faces the sharpest light.
Failure fixes: if one face drifts to a sibling's, regenerate with that person's clearest photo placed first in the upload order. If a child comes back aged up, add "keep each person's apparent age exactly as in their photo" to the lock. And edit only photos of people happy to be edited — for kids, that's the parents' call. Diwali's main night is Sunday, November 8, 2026; generate the portrait the week before. The solo frames pair with it: her diya portrait, his terrace portrait, and the couple rangoli prompt.
How to use this prompt
- Copy it — hit the copy button above to grab the raw prompt text.
- Fill the brackets — replace anything in [BRACKETS] with your own details.
- Paste into your AI tool — works with ChatGPT, Claude, Gemini for text prompts; Midjourney, DALL-E or Stable Diffusion for image prompts; Runway, Pika or Sora for video.
- Iterate — tweak one bracket at a time until the output is exactly right.
Frequently asked questions
How many family members can the prompt handle?
Three to five faces is the reliable range. Beyond five, split into two portraits — per-face quality drops with each added identity.
Can I use one group photo instead of individual photos?
Individual photos work far better, because the prompt names each one. If you only have a group photo, crop each face into its own image first — that prevents most swaps.
Gemini swapped two faces anyway. What now?
Reorder the uploads so the swapped people aren't adjacent, add one distinguishing detail per person to the lock ('the father wears glasses', 'the mother has a nose ring'), and regenerate.
Is this free to use in Gemini?
Yes — a standard Nano Banana edit on the free tier, within daily limits.
