Signed Joke, made with GPT Image 2.5 Sunburst
Example made with GPT Image 2.5 Sunburst

Text to imagePortraits

GPT Image 2.5 Sunburst portrait promptSigned Joke

The exact text-to-image prompt behind this picture, made with GPT Image 2.5 Sunburst.

Prompt

Vertical 4:5 portrait frame. Two adult friends share a joke in sign language against a matte plum community center wall, the nearer friend wearing visible hearing aids, a mint buttoned shirt and cream trousers, the other in a coral sweater. Frame both from the waist up with expressive faces and natural hand gestures clearly visible. Soft frontal daylight, 65mm documentary portrait, realistic skin texture, warm inclusive joy, no text.

Prompts are kept in their original English to help reproduce the example. Copy one as it is, or adapt it for your project.

Opens GPT Image 2.5 Sunburst with this prompt and the settings below filled in.

135 tested GPT Image prompts →

Settings

Mode
Text to image
Aspect ratio
4:5
Quality
high
Price
From 2 credits

The exact price shows before you generate.

Make your own version

  1. The prompt ends with no text. Leave that in, or the model may add random lettering.

Portrait prompt guide

Why this portrait prompt works

Two adult friends laughing in sign language against a matte plum wall, shot as a vertical 4:5 documentary portrait. It works because the prompt fixes the people, their clothes, the framing and the light in short, concrete phrases. Hands, faces and hearing aids are named outright, so the model treats them as the story of the picture.

  1. Format and orientationVertical 4:5 portrait frame.

    This sets the tall shape up front so both people fit from the waist up. If you drop it, the model may pick a wide frame and push the friends apart, leaving less room for the hands.

  2. Subject and settingTwo adult friends share a joke in sign language against a matte plum community center wall

    It gives the action (a shared joke), the setting and a flat plum backdrop. Without 'matte', the wall can pick up shine and glare, and without 'adult' the ages are left to chance.

  3. Identity and outfitsthe nearer friend wearing visible hearing aids, a mint buttoned shirt and cream trousers, the other in a coral sweater

    Naming the hearing aids, the colors and the garments makes the two friends easy to tell apart and keeps the outfits stable. If you cut the colors, the clothes may blend into the plum wall.

  4. Crop and handsFrame both from the waist up with expressive faces and natural hand gestures clearly visible.

    Waist-up framing leaves room for the signing hands and keeps the faces readable. Remove it and the crop may tighten to the faces, cutting the hands or making them small and distorted.

  5. Light, lens and moodSoft frontal daylight, 65mm documentary portrait, realistic skin texture, warm inclusive joy, no text.

    Soft front light keeps faces even, 65mm gives a natural portrait look, and the skin texture phrase avoids a plastic finish. 'No text' stops stray lettering appearing on the wall or the clothes.

Keeping the same face

To keep the same two people across a set, repeat the fixed details word for word and change only the setting or pose. Over-describing new features makes the faces drift.

Visible hearing aids
Keep this phrase unchanged for the nearer friend. It is the clearest identity marker, and a shorter version like 'hearing aids' can lead to them being hidden or dropped.
Mint buttoned shirt and cream trousers
Copy the full outfit exactly in every new prompt. Changing even one color word can alter the shirt style and make the person look different from shot to shot.
Coral sweater on the second friend
Lock this as the second person's label. If you swap in a new outfit, also add a face detail such as hair length so they stay distinct from the first friend.
Your own photo
To use a real person, switch to an image-edit model with that photo as input. Then replace the outfit phrases with 'same person as the input photo' and keep the plum wall and lighting lines.

Three ways to make it yours

Older couple at a library

Use this for older adults with deeper skin detail and wrinkles; the lighting and lens stay the same.

Prompt

Vertical 4:5 portrait frame. Two adult friends in their seventies share a joke in sign language against a matte sage green library wall, the nearer woman wearing visible hearing aids, a navy cardigan and gray trousers, the other man in a mustard sweater. Frame both from the waist up with expressive faces and natural hand gestures clearly visible. Soft frontal daylight, 65mm documentary portrait, realistic skin texture, warm inclusive joy, no text.

Two men in a workshop

Pick this to move the scene to a work setting with a different skin tone and build.

Prompt

Vertical 4:5 portrait frame. Two adult Black men in their thirties share a joke in sign language in a tidy woodworking studio with a matte charcoal wall, the nearer man wearing visible hearing aids, a rust canvas apron over a white tee, the other in a teal flannel shirt. Frame both from the waist up with expressive faces and natural hand gestures clearly visible. Soft frontal daylight, 65mm documentary portrait, realistic skin texture, warm inclusive joy, no text.

Women outdoors in a garden

Choose this for an outdoor feel and more varied body types while keeping the same camera look.

Prompt

Vertical 4:5 portrait frame. Two adult women, one South Asian and one with a fuller figure and short silver hair, share a joke in sign language in front of a matte terracotta garden wall, the nearer woman wearing visible hearing aids and a cream linen shirt, the other in a cobalt blue sweater. Frame both from the waist up with expressive faces and natural hand gestures clearly visible. Soft frontal daylight, 65mm documentary portrait, realistic skin texture, warm inclusive joy, no text.

Model and settings

Keep the 4:5 ratio, because the tall crop suits two people from the waist up and leaves room for the hands. Keep quality on high, since skin texture, hearing aids and finger detail are the points that suffer at lower quality. For a wider group or a banner, change the ratio to a landscape shape and rewrite the framing line to 'to the knees' so nobody is cropped awkwardly. This is a text-to-image prompt, so no input pictures are needed.

  • FLUX.2 ProPick FLUX.2 Pro when you want a strongly photoreal look with exact color control, for example matching the mint, coral and plum to specific hex values.
  • Seedream 5.0 ProPick Seedream 5.0 Pro if you want to guide the look with reference pictures, such as keeping two real friends' faces consistent across a series.

If the portrait comes out wrong

Hands come out with extra or fused fingers, so the signing looks wrong.
Keep 'natural hand gestures clearly visible' and add 'hands held in front of the chest, fingers separated and relaxed'. Regenerate at high quality.
The skin looks smooth and plastic.
Keep 'realistic skin texture' and add 'visible pores, fine laugh lines, no retouching'. Avoid words like 'flawless' or 'perfect'.
The hearing aids vanish or turn into earbuds.
Change the phrase to 'small beige behind-the-ear hearing aids clearly visible on the nearer friend' and make sure the hair does not cover the ears.
Outfit colors shift, for example the mint shirt turns green or blends into the wall.
Repeat each color with its garment, such as 'mint buttoned shirt', and keep 'matte plum wall'. Add 'clear contrast between clothing and wall' if needed.

Questions about this portrait prompt

Can AI draw sign language hands correctly?

It can show believable signing-like gestures, but it will not reliably form real signs. For this prompt, treat the hands as expressive body language, and ask for relaxed, separated fingers to avoid odd shapes.

Why does the prompt say 'no text'?

GPT Image 2.5 Sunburst renders lettering well, so it may add signs or words to a community center wall on its own. 'No text' keeps the plum backdrop clean and the focus on faces and hands.

How do I make the friends look like specific people?

Describing features alone gives new faces each time. To match real people, use an image-edit mode with a clear photo of each person, and keep the outfit, plum wall and soft daylight lines from this prompt.