Google: Nano Banana 2 (Gemini 3.1 Flash Image)

Generate a product photo of a minimalist ceramic coffee mug on a wooden table with soft morning light from the left. Clean background, professional e-commerce style.

The model produces a photorealistic rendering of a matte white ceramic mug centered on light oak planks. Morning sunlight streams from the left, casting a soft shadow to the right and highlighting the mug's subtle texture. The background fades to a clean, slightly warm off-white. The composition follows rule-of-thirds framing with sharp focus on the mug's rim and handle, while the wood grain remains visible but secondary. Color temperature leans warm without oversaturation.

Demonstrates strong product photography fundamentals with accurate lighting physics and material rendering. The 131k token context window allows detailed scene descriptions with multiple lighting cues. However, at $3/Mtok output pricing, iterating on composition details becomes expensive compared to specialized product-shot models.

Create an editorial illustration for a tech article about distributed systems: abstract geometric nodes connected by glowing data streams, isometric view, dark blue and cyan palette.

The model generates an isometric grid of floating hexagonal nodes in varying sizes, connected by animated-looking cyan light trails that curve through three-dimensional space. The background gradients from deep navy at the edges to a lighter blue-gray at the center, creating depth. Each node features subtle internal geometric patterns suggesting circuitry. The overall aesthetic balances technical precision with visual accessibility—clean enough for a business publication, detailed enough for a technical blog.

Handles abstract technical concepts well, translating system architecture metaphors into coherent visual language. The multi-modal input (text + reference images if needed) supports complex briefs. Trade-off: the Flash architecture prioritizes speed over fine detail, so intricate line work or typography within the illustration may lack crispness at high zoom levels.

Design a whimsical character: a small robot librarian with brass gears visible through glass panels, holding a tiny book, Studio Ghibli-inspired soft rendering style.

The model produces a charming knee-high robot with rounded proportions and oversized optical sensors that convey gentle curiosity. Visible brass clockwork gears rotate behind curved glass panels in its torso, rendered with Ghibli's characteristic soft edges and warm lighting. The robot cradles a leather-bound miniature book in articulated copper fingers. Watercolor-like textures blend mechanical precision with organic warmth—panel seams look hand-riveted rather than machined. Background suggests a cozy library corner with defocused warm bokeh.

Shows versatility in stylistic interpretation beyond photorealism, capturing specific artistic influences when prompted. The Gemini foundation handles style transfer requests effectively. Limitation: character consistency across multiple generations isn't guaranteed without additional tooling, making this better for one-off illustrations than character design pipelines requiring exact reproductions.

When Nano Banana 2 handles receipt OCR for small finance teams

A 4-person startup processing 200-400 expense receipts monthly needs accurate line-item extraction without burning budget on vision models. Nano Banana 2 hits the sweet spot here: $0.50/Mtok input means each receipt image costs under a cent to parse, and the 131k context window lets you batch 50+ receipts in a single call with your extraction schema. Output at $3.00/Mtok stays cheap because structured JSON responses are token-light. The model handles standard receipt formats (dates, totals, vendor names) reliably enough for human review workflows. If you're above 2,000 receipts/month or need multi-language invoice parsing, step up to a larger vision model. Below that threshold, Nano Banana 2 keeps your expense automation under $20/month.

Why Nano Banana 2 works for real-time listing moderation at scale

A regional e-commerce platform reviewing 5,000 seller-uploaded product images daily needs fast policy checks (no prohibited items, no misleading angles) before listings go live. Nano Banana 2's $0.50 input pricing makes high-volume moderation feasible: 5k images/day runs about $25/month in API costs. The 131k context lets you include your full policy doc and 10-15 reference images in every call for consistent decisions. Speed matters here—Flash-class models return verdicts in under 2 seconds, keeping seller onboarding smooth. The trade-off: without public benchmarks, you'll need a 500-image validation set to confirm accuracy on your specific prohibited categories. If false negatives cost you more than $500/month in chargebacks, budget for a frontier vision model instead.

When Nano Banana 2 gives clients async design critiques without burning hours

A 6-person branding agency wants to give clients preliminary feedback on logo drafts and mockups between review calls, but designers can't spend 20 minutes per write-up. Nano Banana 2 turns this into a 2-minute task: upload the design, paste the brief into the 131k context, get a 200-word critique covering alignment, readability, and brand consistency. At $0.50 input, each critique costs a few cents—cheap enough to run on every draft iteration. The output quality won't match a senior designer's eye for subtle type hierarchy issues, but it catches obvious problems (low contrast, off-brand colors) and frames them in client-friendly language. This works for agencies doing 10-30 projects monthly. Above that volume, train a fine-tuned model on your past feedback corpus.