Nano Banana Pro: reasoning-powered visuals and accurate text

Unlike pure diffusion models that paint from noise alone, Nano Banana Pro reasons about layout first. That shows up as cleaner proportions, more legible text, and fewer “almost right” iterations on complex scenes.

It supports text-to-image and conversational editing: describe a change in plain language instead of drawing masks manually for many tasks.

Fast generation. Full fidelity.

From fal.ai's Nano Banana 2 page: built on Gemini 3.1 Flash Image with reasoning-guided generation—vibrant, high-fidelity images in about 5–10 seconds with punchy contrast and coherent structure.

Prompt with an art-director brief: subject, action, setting, style, camera, light, quoted text, and constraints. See the full Nano Banana Pro prompting guide.

Text rendering

fal.ai highlights legible, correctly spelled text in generated images—signage, labels, watermarks, and packaging with character-by-character validation.

Prompt

Copy

A hyperrealistic product photo of a matte ceramic coffee mug on a wooden table, morning light streaming through a window casting soft shadows, steam rising from the cup with the text 'GOOD MORNING' embossed on the side in a clean serif font, shallow depth of field, 85mm lens

Edit with words, not masks

Describe changes in plain language—replace backgrounds, change colors, add or remove objects, adjust lighting—without manual selection. Supports up to 14 reference images and multi-image compositing on fal.ai.

Nano Banana 2.1 is newer in Google's Flash line with mask editing and token billing on many platforms. Nano Banana Pro remains the choice when typography and multi-step reasoning dominate. For a deep dive on Google's latest Flash model, visit Nano Banana 2.1.

Nature and landscape imagery

Prompt

Copy

An aerial view of a bioluminescent bay at night, glowing blue-green plankton tracing the wake of a small wooden boat, stars reflected in the calm water, long exposure effect, the Milky Way arching overhead, National Geographic cover quality

Campaign series where the same character appears in three environments, quote wardrobe and hair so the model locks identity.

Poster and packaging mockups with long headlines; put exact strings in quotation marks and specify type style (serif, condensed sans, all caps).

Infographic-style frames with labeled regions; use thinking-heavy settings on hosts that expose them when small type must stay legible.

Product photography with text rendering

Prompt

Copy

A hyperrealistic product photo of a matte ceramic coffee mug on a wooden table, morning light streaming through a window casting soft shadows, steam rising from the cup with the text 'GOOD MORNING' embossed on the side in a clean serif font, shallow depth of field, 85mm lens

Cinematic portrait photography

Copy any prompt below and try it in the fal playground—verbatim from their model page.

Prompt

Copy

A cinematic portrait of an elderly Japanese fisherman mending nets at dawn, weathered hands in sharp focus, fishing village harbor in soft bokeh behind him, golden hour side lighting, documentary photography style, Leica 50mm Summilux wide open

Conversational edits shine when you change one variable: “keep layout, swap dusk for noon,” “same pose, replace jacket with raincoat.”

When a host supports multiple references, bring brand color swatches or logo plates as images instead of describing hex codes in prose alone.

Cross-check mask-friendly edits on Nano Banana 2.1 if you need region-locked changes at Flash-class latency.

Architectural visualization

Prompt

Copy

A detailed architectural visualization of a modern hillside house with floor-to-ceiling glass walls, infinity pool merging with the ocean horizon, warm interior lighting at dusk, tropical vegetation, rendered in photorealistic style with accurate reflections and volumetric light

UltraPix runs generation on Android with credits you can try before subscribing. Start with our how to use Nano Banana Pro–style workflows article, then get the app.

Paste structured briefs from the blog, generate, then chain background removal or upscale without leaving the app gallery.

See what Nano Banana 2 can create

Same architectural brief as fal.ai's playground gallery—useful for Pro-class prompt structure (quoted type, materials, and light) in mobile apps like UltraPix.

Prompt

Copy

A detailed architectural visualization of a modern hillside house with floor-to-ceiling glass walls, infinity pool merging with the ocean horizon, warm interior lighting at dusk, tropical vegetation, rendered in photorealistic style with accurate reflections and volumetric light

Google surfaces Pro-class image generation inside Gemini consumer apps, Search AI Mode, and developer APIs. Third-party creative platforms re-host the same weights with their own credit meters.

Token pricing on APIs often scales with output resolution and reference count more than prompt length, budget 4K campaign finals separately from chatty exploration at 1K.

When a host exposes thinking levels, raise them for dense layouts with small type; keep medium for portrait and product work to save latency.

Always verify model IDs in the dashboard you pay, marketing names change faster than blog posts.

Quote display type and label copy exactly.

Specify thinking level when small type fails.

Reuse character references across a series before changing wardrobe.

One edit instruction per pass for conversational tools.

Compare 1K drafts before spending 4K credits.

Paste winners into UltraPix on Android for cutout and upscale follow-ups.