Describe the viewpoint, light and materials to create a photographic image, and use it as a concept rather than presenting it as a real photograph.
Try an example
The same breakfast-table prompt, run twice. Only the camera distance changed.
Both versions share the ceramic plate, one pastry, the coffee cup, a small window table and the warm early-morning mood, in the same 16:9 frame. Small details can still differ between runs, because generation is not pixel-for-pixel repeatable.
Only the camera distance: a close view that emphasizes texture and steam, then a wider view with room for the setting and for editable promotional copy on the left. Compare which one fits your banner layout.
Inspect cup handles, cutlery touching the plate, pastry layers, window reflections and shadow direction at full size. Check that the quiet area on the left survived. Label the result as a concept: it is not a photograph of an item that is served.
Generated photos suit planning and illustration, where a photographic finish helps. Each card names the job and what to verify.
16:9 · Article header
Some ideas are hard to photograph literally. For an article about starting a small creative business, describe a modest worktable with paper samples, a lamp and morning light, and leave an uncluttered area for the title. Check the likely mobile crop, and inspect any notebook or screen for lettering that looks real from a distance but fails up close.
3:4 · Product concept
Compare a pale stone surface in soft window light with a darker shelf and a stronger shadow before you book a real shoot. Describe the arrangement separately from the product’s identity: where the object sits, which materials surround it and how much empty space the layout needs. The concept guides a discussion; it does not verify packaging or ingredients.
3:2 · Location mood
Describe the qualities you need, such as a narrow cafe with warm wood or a workshop with large windows, instead of naming a real street. Specify architecture, available space, light direction and foreground against background. Label each image so collaborators do not assume a room is booked, and compare the board with real locations once they exist.
Three steps in the creation card above.
Write the viewpoint, light and materials of a fictional scene in the prompt field, up to 4,000 characters, and say where the layout needs quiet space. Name a place by its qualities, such as architecture and light direction, rather than by a real street. If exact wording matters, add it later in your own layout.
Add a reference only when it answers one question: the subject, a color palette or a composition, and say which in the prompt. Then choose the model, aspect ratio, HD, 2K or 4K, and one, two or four outputs. Upload only material you have the right to use; a reference does not guarantee a face, logo or texture stays unchanged.
One Nano Banana HD image costs 23 credits and one Nano Banana Pro HD image costs 45, shown on Generate. Change one decision per attempt, such as light direction or camera angle, and settle the composition before paying for a higher resolution, which adds pixels but not accuracy.
Four photographic choices you set in the prompt, each one visible in the result.
01
An AI photo generator needs the decisions a photographer makes first: how close, how high and how wide. One plate seen close up and a whole breakfast table seen from above are different pictures of the same subject. Say the framing in the prompt, then compare two distances side by side.
02
Window light in the early morning, hard midday sun and a soft overcast sky each give a scene a different mood and different shadows. Name the direction and the warmth of the light, and the AI photo generator builds the surfaces around it. Check that shadows fall consistently across the whole frame.
03
Glazed ceramic, brushed metal, linen and wood read as photographic when the prompt says what each surface is. Detail comes from naming a few materials rather than many. Zoom to full size and check edges, reflections and small objects before you treat the picture as more than a concept.
04
A 16:9 header, a 3:4 product concept and a 3:2 location mood frame need different compositions. Choose the aspect ratio first, then the resolution and the file format. The quote on Generate updates with every change.
Pick the page that matches the kind of image you need next.
An AI photo generator, sometimes called an AI photo maker, makes a photographic-looking image from a prompt and optional visual references. The image can use realistic light, framing, and surface detail. Potto AI creates a generated raster output for review; the photographic appearance does not prove that the scene, person, or item existed.
No. A real photograph records light from a scene through a camera; a generated photo-like image is synthesized from a request. Use real photography when documentation, product accuracy, or event evidence is the task. Label generated concepts when their origin matters to the audience.
Yes, as a visual planning exercise rather than product evidence. Compare camera angle, backdrop, light direction, and space for copy. Check the actual product’s details separately, and record which aspects of the concept the real production should reproduce. The generated scene does not replace a verified photograph of the sold item.
Yes, as planning or illustration. A prompt that names a city can produce a plausible scene without reproducing a real street, so describe architecture, space, light direction and viewpoint instead, and label the board so nobody assumes a location is booked. For a restaurant, verify any public image against the actual dish and portion; the generator cannot confirm ingredients, serving quantities or nutritional information.
Yes, for fiction or a visual exercise. Describe an original adult character through visible choices such as posture, clothing and setting, and avoid naming a real person. Do not present the result as a real customer, employee, expert or testimonial subject. If you want a portrait of yourself from an authorized photo, start from a headshot template instead.
Change one visual decision at a time, such as light direction or camera angle, and keep the rest of the brief steady. Save the wording of the accepted direction before experimenting further. Higher resolution adds pixels but does not make anatomy, labels or reflections more accurate, so settle the composition first, then choose the output size for the destination.
You can describe the photo, add references and pick settings before signing in. Generate opens the sign-in dialog, and your prompt and settings are kept.
The task shows why it stopped and offers a retry, and the reserved credits are released. Cancel a running request from its controls; leaving the page does not stop it. A new photo after a finished result is a new, charged request.
Each photo costs 45 credits with Nano Banana Pro at HD or 2K, or 23 with Nano Banana at HD; 4K and more photos cost more. Generate shows the total before you confirm.
Describe viewpoint, light and materials, then check the result before you use it as anything more than a concept.