Nano Banana Pro API: Gemini 3 Pro Image, Specs and Pricing
Nano Banana Pro is Google Gemini 3 Pro Image. Photoreal output, legible in-image text, camera and lighting controls, up to 20 references — $0.06 per image.
Nano Banana Pro is Google's production-tier image model — Gemini 3 Pro Image — and the third member of the Nano Banana family on Unifically. It generates photoreal output, renders legible in-image text across multiple languages, and exposes camera and lighting controls that the other Nano Banana variants do not. Released November 20, 2025, it is the first Nano Banana that ships ready-to-use assets without a retouching pass. On Unifically it is $0.06 per image with the same prompt-and-references surface as Nano Banana 2.
TL;DR: Nano Banana Pro is Gemini 3 Pro Image. It accepts a prompt, up to 20 reference images (10 MB max each), and ten aspect ratios from 1:1 through 21:9, with selectable 1K, 2K, or 4K output. Pro is $0.06 per image on Unifically — same flat rate regardless of resolution. Use Pro when you need legible text inside the image, photoreal materials, or coherent multi-subject scenes; use Nano Banana 2 when you are iterating quickly and do not need either.
What is Nano Banana Pro?
Nano Banana Pro is Google's production tier in the Nano Banana family, built on the Gemini 3 Pro Image model. Unlike the base Nano Banana (Gemini 2.5 Flash Image) and Nano Banana 2 (resolution-tiered Gemini 2.5 Flash Image), Pro runs an explicit reasoning step before drawing. That reasoning step is what makes the model qualitatively different in three places: in-image text, photoreal materials, and multi-subject composition.
It is the same simple API surface as the rest of the family — prompt, optional reference images, aspect ratio, resolution. The differences are inside the model:
- Photoreal output — material rendering on metal, glass, fabric, and skin holds up at print resolution.
- Legible in-image text — multi-line copy renders correctly in English, Japanese, Korean, Arabic, and more. Other Nano Banana variants distort anything past three or four words.
- Multi-subject coherence — the reasoning pass keeps complex scenes (six-plus elements) composed instead of dropping or merging subjects.
- Camera and lighting controls — explicit angle, depth of field, bokeh, and lighting language work as prompt controls, not just stylistic hints.
Every output ships with Google's invisible SynthID watermark for AI-content provenance.
Why Nano Banana Pro matters in 2026
Two things kept earlier text-to-image models out of ship-ready production work, and Pro is the first Gemini-line model to fix both at once:
- In-image typography. Marketing posters, infographic callouts, and localized campaign assets all need legible body copy inside the rendered image. Most image models fail here past a single short phrase — Pro renders multi-line type correctly across multiple writing systems.
- Material believability. Product hero shots and editorial photography need plausible reflections, refractions, and surface detail. Pro produces output that survives a scroll past a real photograph; Nano Banana 2 still reads as AI-generated on close inspection.
Combined, those two unlock a class of work that previously required either a retoucher or a stylist. The cost of that quality is speed: the reasoning pass adds latency. Pro is the right pick for ship-grade assets, not for fast iteration loops.
How Nano Banana Pro works
Inputs
- Prompt — required text describing the image. Camera and lighting language (e.g. "85mm portrait, soft window light, shallow depth of field") works as a real control.
resolution—1k,2k, or4k(selectable output, flat per-image price).aspect_ratio— choose from ten ratios spanning 1:1 through 21:9.image_urls— up to 20 optional reference images (10 MB max each). Used for character consistency, style transfer, or multi-image fusion.
Output
- A single PNG at the chosen resolution and aspect ratio.
- SynthID invisible watermark embedded for AI provenance.
- Returned via task-based async polling.
Nano Banana Pro pricing and how it compares
| Source | Variant | Resolution | Per-image price |
|---|---|---|---|
| Unifically | Nano Banana Pro | 1K, 2K, or 4K | $0.06 |
| Unifically | Nano Banana | single tier | $0.03 |
| Unifically | Nano Banana 2 1K | up to 1K | $0.03 |
| Unifically | Nano Banana 2 2K | up to 2K | $0.05 |
| Unifically | Nano Banana 2 4K | up to 4K | $0.06 |
| Google direct | Gemini 3 Pro Image | varies by tokens | per Vertex / Gemini API rate card |
Pro is a flat rate regardless of resolution, which is the main pricing difference from Nano Banana 2: with Pro you do not pay more for 4K. At 4K specifically Pro and Nano Banana 2 both list at $0.06, and the choice comes down to whether you need Pro's reasoning, text rendering, and camera controls.
Nano Banana Pro vs Nano Banana vs Nano Banana 2
Three Nano Banana models ship on Unifically. Pick by what you actually need:
| Variant | Underlying model | Resolution | Best for |
|---|---|---|---|
| Nano Banana | Gemini 2.5 Flash Image | single tier | Cheapest list price; fast drafts; high-volume social |
| Nano Banana 2 | Gemini 2.5 Flash Image | 1K, 2K, 4K | Draft-to-final workflows on a single prompt and reference set |
| Nano Banana Pro | Gemini 3 Pro Image | 1K, 2K, or 4K | Studio-grade work needing legible text, photoreal materials, and multi-subject scenes |
Both Nano Banana 2 and Nano Banana Pro reach 4K. The split is qualitative: Pro is the only one with a reasoning pass, in-image text rendering, photoreal material output, and explicit camera and lighting controls. Pro is also slower per call because of the reasoning step — for fast iteration, draft on Nano Banana 2 and promote selected takes to Pro for the final render.
How to call Nano Banana Pro
The API is async: POST a generation, then poll until the image URL is ready.
const API = 'https://api.unifically.com';
const headers = {
Authorization: `Bearer ${process.env.UNIFICALLY_API_KEY}`,
'Content-Type': 'application/json',
};
const start = await fetch(`${API}/nano-banana-pro/generate`, {
method: 'POST',
headers,
body: JSON.stringify({
prompt:
'Editorial poster for a Tokyo coffee festival. Top half: a single espresso cup on a brushed-steel counter, 85mm shot, soft window light, shallow depth of field. Bottom half: clean two-line headline in Japanese — "東京コーヒー祭 / 2026" — set in a modern sans-serif',
resolution: '4k',
aspect_ratio: '4:5',
image_urls: ['https://example.com/brand-mark.png'],
}),
}).then((r) => r.json());
while (true) {
await new Promise((r) => setTimeout(r, 2000));
const task = await fetch(`${API}/v1/tasks/${start.task_id}`, { headers }).then((r) => r.json());
if (task.status === 'completed') {
console.log(task.image_url);
break;
}
if (task.status === 'failed') throw new Error(task.error);
}
The same payload shape works for all Nano Banana variants — switch the path to /nano-banana/generate or /nano-banana-2/generate to call the other tiers without changing your client code.
Common mistakes and gotchas
- Defaulting to Pro for everything. Pro is twice the cost of base Nano Banana at 1K and slower per call. Draft on Nano Banana or Nano Banana 2 1K, promote to Pro only on takes that survive creative review.
- Treating Pro and Nano Banana 2 4K as interchangeable. Both list at $0.06. They are not the same model. Pro uses Gemini 3 Pro Image with a reasoning pass; Nano Banana 2 4K uses Gemini 2.5 Flash Image without one. If your output needs legible text or photoreal materials, the price match is misleading — only Pro delivers them.
- Stuffing 20 references when you only need 2. Reference images compete for influence. Over-loading dilutes the strongest signal. Send the minimum that anchors the look, especially for character-consistency work.
- Vague camera language. Pro's camera controls are real prompt levers. "85mm portrait, soft window light, shallow depth of field" does something specific; "professional photo" does not. Treat camera and lighting language like Photoshop sliders.
- Asking for paragraphs of body copy. Pro renders multi-line text correctly, but text generation is not a typography engine. For long-form copy, generate the imagery and headline in Pro and composite the body copy in your design tool.
- Stripping SynthID from outputs. The invisible watermark is part of Google's AI provenance signal. Pipelines that strip metadata also strip the SynthID flag — keep it for compliance.
Frequently asked questions
What is Nano Banana Pro?
Nano Banana Pro is Google's Gemini 3 Pro Image model, the production tier of the Nano Banana family. It accepts a prompt and up to 20 reference images, runs an explicit reasoning step before drawing, and produces photoreal output with legible in-image text and explicit camera and lighting controls. It went generally available on November 20, 2025.
How much does Nano Banana Pro cost?
$0.06 per image on Unifically, flat across 1K, 2K, and 4K resolution. The base Nano Banana variant is $0.03; Nano Banana 2 is tiered at $0.03 / $0.05 / $0.06 for 1K / 2K / 4K. Google direct prices Gemini 3 Pro Image per the Vertex / Gemini API rate card.
How is Nano Banana Pro different from Nano Banana 2?
Both run on Gemini and both reach 4K with the same ten aspect ratios and same 20-reference cap. Pro swaps in Gemini 3 Pro Image, adds a reasoning pass that holds multi-subject scenes together, renders in-image text legibly across English, Japanese, Korean, Arabic, and more, and exposes camera and lighting controls. Nano Banana 2 stays on the original Gemini image lineage, skips the reasoning step, and is faster per call but cannot produce legible text or photoreal materials. Pick Pro for ship-ready assets, Nano Banana 2 for fast iteration.
How many reference images can I send to Nano Banana Pro?
Up to 20 reference images per request, each up to 10 MB. Send fewer references to give each one more influence — over-loading dilutes the strongest one. For character consistency, two or three high-quality references usually outperform twenty mixed ones.
Does Nano Banana Pro support text in non-English languages?
Yes. Multi-line in-image text renders correctly across English, Japanese, Korean, Arabic, and more. This is one of the model's signature capabilities and the main reason to pick Pro over Nano Banana 2 for localized campaign work.
Is Nano Banana Pro slower than Nano Banana 2?
Yes. Pro runs a reasoning step before drawing, which adds latency per call. The trade is quality and prompt adherence on harder scenes. For high-volume drafting, use Nano Banana 2; for final assets, Pro is worth the wait.
Does Nano Banana Pro include a watermark?
Yes. Every output ships with Google's SynthID invisible watermark for AI-content provenance. The watermark is not visible to viewers but can be detected by Google's verification tooling and AI-content classifiers.
Related reading
- Nano Banana Pro model page — live playground at the $0.06 tier
- Nano Banana — base $0.03 Gemini 2.5 Flash Image variant
- Nano Banana 2 — resolution-tiered Gemini image generation at 1K, 2K, or 4K
- GPT Image 2 and Flux.2 — alternative image APIs on Unifically



