AI Character Photo Ideas: 18 Scenes and How to Ask for Them

How do you describe a scene so the image comes out well?

An image model fills every gap you leave with a guess. If you write “a woman at a cafe”, it decides the light, the angle, the mood and the pose for you, and the result is often generic. The fewer decisions you leave open, the closer the picture gets to what you had in mind.

You do not need a long prompt. One short line that covers five points is usually enough, and you can write it the way you would tell a friend.

• Setting: where the character is, such as a rooftop, a library or a train.
• Time of day and light: soft morning light, golden hour, a desk lamp, neon at night.
• Framing: close-up on the face, half-body, or full-body so the place is visible.
• Mood: calm, playful, tired in a nice way, curious.
• Activity: reading, walking, cooking, looking out of a window.

What is the difference between a weak and a strong request?

Compare “a girl in a kitchen” with “small evening kitchen, warm yellow light, half-body, stirring a pot, relaxed and smiling”. The second one fixes the place, the light, the framing, the mood and the action in one breath. The first leaves all of them to chance.

Pick one or two details that carry the picture rather than listing twenty. Light and framing do the most work, because they change how the whole image feels. Add small objects only when they matter, such as a steaming cup or a rain-covered window.

If you are stuck, start with the light. “Rainy grey morning” or “sunset from behind” already implies a place and a mood.

Which scenes are worth trying first?

Here are eighteen everyday scenes that give a model plenty to work with. Each one names a place, a light and an activity, so you can use it as is or change a detail. Mix framings between close-up and half-body to see which ones you like.

• Morning coffee by a rainy window, soft grey light, close-up.
• A rooftop at sunset, warm orange light, half-body, wind in her hair.
• A quiet library between tall shelves, lamp light, reading.
• A train at night, head resting on the window, city lights passing.
• A bakery just after opening, flour in the air, warm yellow light.
• A park bench in autumn, a scarf, fallen leaves, early afternoon.
• A seaside promenade at dawn, cool blue light, full-body, walking.
• A small kitchen in the evening, steam from a pot, half-body.
• A record shop, flipping through vinyl, soft overhead light.
• A balcony full of plants at golden hour, a cup of tea.
• A night market with paper lanterns, glowing colors, looking around.
• A snowy street, a hot drink, breath visible in the cold.
• A bookshop cafe, writing in a notebook, window light.
• A mountain trail lookout at midday, a backpack, a wide smile.
• An art studio with paint on her hands, afternoon light.
• A late-night diner booth, a neon sign, a milkshake.
• A spring garden, watering flowers, soft morning sun.
• A sofa under a blanket with a film on, the glow of the screen.

Automatic photos or the Show me a moment button?

In CraveMe Private there are two ways to get an image. By default the bot generates one automatically every third message, based on what you are talking about right then. If you chat about a rainy evening, the next picture will probably show it. You can turn automatic images off in settings.

The second way is the “Show me a moment” button under every character message. You describe the scene you want, using the five points above, and the image arrives in a few seconds. Automatic photos are good for surprise and for keeping the conversation visual. The button is better when you have a specific scene in mind.

Each generated image costs one credit, and you get 8 free credits when you start. Credits can be bought with Telegram Stars if you want more.

What makes the same face carry across scenes?

A scene prompt describes where and how, not who. If the face were left to the model, it would invent a new one for every picture. In CraveMe you create your character once: you upload a portrait you have the right to use as inspiration, and the app generates a new, original face that becomes the permanent reference.

Every later image gets a new scene, composition and pose, with that same generated face applied. So you can ask for a rooftop, a library and a night train and still recognize one character. You describe the scene and do not need to repeat the face.

Faces still read differently under heavy shadow, far-away framing or dramatic light, so expect small variations. Half-body and close-up shots show the face most clearly.

What common mistakes spoil an image?

The most common one is stacking too many ideas into one request: three locations, two outfits and four moods. The model blends them and the picture looks confused. Keep one place, one light and one action per photo, and ask for the next idea in a new request.

Another is leaving out framing. Without it you may get a distant figure when you wanted a face, or the reverse. Name the framing every time. Finally, keep things everyday and clear. Simple scenes with strong light usually look better than complicated ones.

FAQ