Nano Banana 2 prompts work best when they lock down subject, lighting, camera, and aspect ratio in one pass. This guide gives you copy-paste examples and a repeatable structure you can reuse across portraits, headshots, style transfers, and 3D collectible figures.
What makes Nano Banana 2 prompts different
Nano Banana 2 (often written as nano banana nano banana 2) responds like a photo editor that also understands scene direction. It holds identity across edits, follows spatial instructions (left/right, foreground/background), and respects camera language. That means your prompt should read less like a wish list and more like a shot brief.
A useful Nano Banana 2 prompt has five parts:
- Subject + action — who or what, doing what, with what expression.
- Environment — location, time of day, weather, background elements.
- Lighting — source, direction, quality, color temperature.
- Camera — lens, aperture, distance, angle, film stock or sensor look.
- Output spec — aspect ratio, resolution intent, and any hard constraints (no text, no watermark).
If you only remember one rule: put the camera and lighting after the subject, not before. Models weight earlier tokens more heavily, and subject drift is the most common failure.
For a deeper breakdown of ordering and syntax, see prompt structure and syntax.
A reusable Nano Banana 2 prompt template
Start from this and swap the bracketed fields:
"[Subject and action], [environment with 2–3 concrete details], lit by [light source + direction + quality], shot on [camera/lens] at [aperture], [angle and distance], [color grade or film look], [aspect ratio]."
Example filled in:
"A ceramicist shaping a bowl on a kick wheel, sunlit studio with clay dust in the air and shelves of bisque-fired pots behind her, lit by a single north-facing window from camera left, shot on a 50mm lens at f/2, eye-level medium shot, warm neutral grade, 3:2."
That prompt gives the model four independent decisions to make (subject, space, light, optics) instead of one vague one. It also gives you four dials to turn when a result misses.
Nano Banana 2 prompt examples you can copy
Portrait with controlled lighting
"Close-up portrait of a 30-year-old man with a short beard and a faded scar over his left eyebrow, plain charcoal backdrop, lit by a softbox from 45 degrees camera right with a subtle rim light from behind, shot on an 85mm lens at f/1.8, chest-up framing, natural skin texture with visible pores, 4:5."
The scar is a specificity anchor — it forces the model to commit to a face rather than averaging one. For more on this, see portrait lighting techniques and skin texture and makeup.
Professional headshot
"Corporate headshot of a woman in a navy blazer, neutral gray seamless background, lit by a two-light setup: key light at 45 degrees camera left, fill light at 1/2 power camera right, shot on a 105mm lens at f/4, head-and-shoulders framing, slight smile, sharp focus on the eyes, 1:1."
Headshots live or die on even lighting and a clean background. Keep the background description to one clause so the model does not invent props. More variations: professional headshots.
Style transfer without losing the face
"Transform the attached photo into a 1970s Kodachrome film still: warm highlights, slightly crushed blacks, visible grain, soft halation around bright edges. Keep the subject's facial features, hairstyle, and expression unchanged. 3:2."
The phrase "keep the subject's facial features, hairstyle, and expression unchanged" is doing real work here. Without it, style prompts tend to redraw the face. Related: vintage retro photography and cinematic film look.
3D collectible figure
"A 3D collectible figure of a samurai in a dynamic mid-swing pose, standing on a circular black base with a small nameplate, photographed on a white studio background with soft top-down lighting, shallow depth of field, product photography style, 1:1."
This is one of the most reliable Nano Banana 2 use cases because the model has seen thousands of product shots of toys. More: 3D action figure prompts.
Environmental portrait with weather
"Environmental portrait of a fisherman mending a net on a wooden dock, overcast morning with light rain, wet planks reflecting the gray sky, lit by flat diffused daylight, shot on a 35mm lens at f/2.8, wide medium shot, muted cool grade, 16:9."
Weather adds texture that studio lighting cannot fake. For more atmosphere control, see weather and environmental effects.
Product shot with reflection
"A matte black wireless earbud case on a polished dark stone surface, single softbox from above creating a soft gradient reflection, shot on a 100mm macro lens at f/8, three-quarter angle, deep shadows, 4:5."
Product prompts benefit from naming the surface material. "Polished dark stone" produces a different reflection than "glass" or "brushed metal." More: product photography prompts.
Building your own Nano Banana 2 prompt library
A prompt library is not a folder of text files. It is a set of reusable modules you recombine. Build it in four layers:
Layer 1 — Subject blocks. Write 10–20 subject descriptions you reuse: "a 40-year-old woman with curly auburn hair and freckles," "a weathered fishing boat with peeling blue paint." Specific, not generic.
Layer 2 — Lighting blocks. Write 8–12 lighting recipes: "softbox at 45 degrees camera right," "golden hour backlight with lens flare," "single overhead fluorescent with green cast." These are the highest-leverage modules because lighting drives mood more than subject does.
Layer 3 — Camera blocks. Write 6–10: "85mm at f/1.8, chest-up," "24mm at f/8, low angle wide," "macro 100mm at f/11." Camera language is what separates a snapshot from a photograph.
Layer 4 — Output specs. Aspect ratios, resolution notes, and negative constraints. Keep a standard negative list: "no text, no watermark, no extra fingers, no logos." For more on this, see negative prompts guide.
Once you have these four layers, a new prompt is a 30-second assembly job. You stop rewriting from scratch and start tuning one variable at a time.
Nano Banana 2 prompt patterns that solve specific problems
Fixing identity drift across a series
When you generate multiple images of the same person, add a fixed identity clause at the start of every prompt: "The same woman as before: oval face, dark brown eyes, small mole on the right cheek, shoulder-length black hair." Repeat it verbatim. Do not paraphrase between prompts.
Controlling composition when the model crowds the frame
Add explicit spatial language: "subject occupies the left third of the frame, negative space on the right." Nano Banana 2 respects left/right instructions more reliably than older models. See image composition techniques.
Getting consistent color across a set
Name a film stock or a grade instead of adjectives. "Kodak Portra 400" or "teal-and-orange blockbuster grade" produces more consistent results than "warm and cinematic." More: color theory for AI images.
Making style transfers keep the original face
Always include three clauses: keep facial features, keep hairstyle, keep expression. Then describe the style. Order matters — constraints first, style second.
Generating believable 3D figures
Add "product photography style" and a base or stand. The base is the tell: real collectibles sit on something. More: 3D render and CGI style.
Common mistakes with Nano Banana 2 prompts
Stacking too many styles. "Cyberpunk watercolor art nouveau with steampunk elements" produces mud. Pick one primary style and one modifier. If you want fusion, do it in two passes. See style fusion techniques.
Describing emotions instead of expressions. "Sad" is vague. "Eyes downcast, mouth slightly downturned, brows drawn together" is specific. See portrait expressions and emotions.
Forgetting aspect ratio. Without it, the model picks a default that may crop your composition. State it every time.
Using vague lighting words. "Good lighting" means nothing. "Soft key light from camera left at 45 degrees" means something.
Ignoring the negative list. If you do not exclude text and watermarks, some outputs will include them. Keep a standard negative block.
Workflow: from prompt to finished image
- Write the prompt using the five-part structure.
- Generate three variations at low effort.
- Pick the closest one and edit one variable at a time — usually lighting or camera.
- Once the composition is right, refine skin, texture, or material details.
- Run a post-processing pass if needed: image post-processing.
This loop is faster than rewriting the whole prompt. It also teaches you which tokens actually move the output.
Where NanoPrompts fits
NanoPrompts is a library of ready-to-use AI image-generation prompts and templates. It provides copy-paste prompts for photo transformations such as portraits, headshots, style transfers, and 3D collectible figures. It focuses on AI image prompts for Gemini / nano-banana style models, not text or writing prompts.
If you want to skip the module-building step, start from a template and swap the subject and lighting blocks. That is the fastest path to a usable result.
Try it
Browse the NanoPrompts library, copy a prompt that matches your subject, and swap in your own details. The examples above are a starting point — the library gives you the rest.
