Nano Banana Pro accepts up to 14 reference images in one Gemini API request. That headline number needs context: Google sets lower high-fidelity limits for objects, characters, and styles, and consumer interfaces can expose different upload controls.
If you want that 14-image workflow without setting up the API yourself, GlobalGPT provides Nano Banana Pro inside an all-in-one AI workspace. You can attach up to 14 references, then use leading models in the same workspace for research, writing, reasoning, and other everyday AI tasks. Image generation uses credits, so the number of generations you can run depends on your plan and remaining balance.
The Short Answer: Up to 14 Reference Images
Google’s Gemini API documentation says Gemini 3 image models can mix up to 14 reference images. Nano Banana Pro is the premium Gemini 3 Pro Image model, identified in the API as gemini-3-pro-image.
The word Referenz matters. These images guide a new composition or edit. The allowance does not mean the model generates 14 outputs, nor does it describe how many generations a free or paid plan includes.
Why 14 Images Does Not Mean Equal Fidelity for Every Reference
Google separates the overall input count from more demanding consistency tasks. For Nano Banana Pro, the documentation lists up to six object images with high fidelity, five character images for character consistency, and three style-reference images. Those categories sit within the broader 14-image total.
The limits inside the 14-image total
Maximum reference images per request
Quelle: Google Gemini API documentation · Checked July 28, 2026
What this means: 14 is the attachment ceiling, not a promise of equal precision across 14 identity-critical inputs. If consistency matters, design the prompt around the smaller category limits.
| Limit type | Maximal | Practical meaning |
|---|---|---|
| Total references | 14 | All references combined |
| High-fidelity objects | 6 | Objects whose defining details must remain recognizable |
| Konsistenz der Zeichen | 5 | People or characters whose identity must remain consistent |
| Style references | 3 | Images used mainly to guide visual style |
Nano Banana Upload Limits by Platform
The model limit and a product’s upload control are different constraints. Use the number for the exact interface where you are working.
Upload controls across four entry points
Maximum files or reference slots per prompt
Quellen: Google Gemini API documentation; Gemini Apps Help; GlobalGPT Nano Banana Pro and Nano Banana 2 interfaces · Checked July 28, 2026
What this means: the consumer Gemini app has the lower attachment ceiling here. The API and both GlobalGPT image-model entries expose 14, but generations on GlobalGPT still depend on available credits.
| Umwelt | Checked allowance | Condition |
|---|---|---|
| Gemini API / Nano Banana Pro | 14 references | Lower fidelity sub-limits apply |
| Gemini-Apps | 10 supported files | Subject to availability |
| GlobalGPT Nano Banana Pro | 14 reference slots | Generation uses credits |
| GlobalGPT Nano Banana 2 | 14 reference slots | Separate entry; generation uses credits |
GlobalGPT Has Separate Nano Banana Pro and Nano Banana 2 Entries
GlobalGPT provides separate entries for Nano Banana Pro and Nano Banana 2. In the interfaces checked on July 28, 2026, both reference panels displayed 0/14, indicating room for up to 14 reference images.
The reference allowance is not a free-generation allowance. Image generation consumes credits, so the number of generations depends on the plan and remaining balance. Check the live Preisseite von GlobalGPT before relying on a fixed credit figure.
How to Get Better Results With Multiple Reference Images
- Start with the fewest useful images. Three to five focused references are easier to control than 14 competing inputs.
- Give every image a role. Identify which reference controls the person, product, background, pose, palette, or typography.
- Prioritize identity-critical references. Use clear angles and adequate resolution for faces, branded objects, or products.
- Avoid near-duplicates. Similar references can add noise without supplying useful new information.
- Split complex work into stages. Establish the subject and composition first, then refine details in a follow-up edit.
For a full workflow, see Wie man Nano Banana Pro verwendet. Developers can review the Nano Banana Pro API-Handbuch, while buyers can compare access in the Preisübersicht für Nano Banana Pro.
Häufig gestellte Fragen
Can Nano Banana Pro take 14 images at once?
Yes. Google’s Gemini API documentation states that Nano Banana Pro can accept up to 14 reference images in one request. High-fidelity limits for objects, characters, and style references are lower.
Can all 14 images contain different people?
You can remain within the 14-image total, but Google documents character consistency for up to five character images. Do not assume 14 different people will all retain equally strong identity consistency.
Does the Gemini App also allow 14 uploads?
No. The Gemini Apps Help page lists up to 10 supported files in the same prompt, subject to availability. That app-level limit is separate from the Gemini API model limit.
How many reference images can GlobalGPT accept?
In the interfaces checked on July 28, 2026, both GlobalGPT’s Nano Banana Pro and Nano Banana 2 entries displayed 14 reference slots. Generating an image consumes credits.
Is the upload limit the same as the generation limit?
No. The upload limit describes references attached to one request. A generation quota or credit balance describes how many requests you can run. Output count and resolution are separate limits.
Endgültiges Urteil
Nano Banana Pro’s official total is up to 14 reference images per Gemini API request. For dependable results, work within the smaller fidelity limits and check the interface you use. Gemini Apps lists 10 supported files, while GlobalGPT’s separate Nano Banana Pro and Nano Banana 2 entries both provide 14 reference slots and use credits for generation.
