Nano Banana Pro - Generador de imágenes de Google AI

Sobre Nano Banana Pro

Nano Banana Pro is the higher-fidelity pass on Google's Nano Banana image model. This page runs it in two modes, text to image and image to image, and accepts up to five reference images so a product or character stays recognisable while the scene changes. Each run costs 240 credits, and the button states the charge before you start.

The base Nano Banana model on this site exists to help you decide what an image should be. Pro is for the frame you keep. It spends more compute on the same prompt, and the difference shows in the parts viewers inspect at full size: product edges, surface texture, reflections, and the way light falls across a face. Both modes use the same price, so moving from a written prompt to reference-led work does not change the cost.

Image to image accepts up to five references in one request, with PNG, JPG, JPEG and WEBP files up to 10 MB each. Write the prompt as an instruction about what should change, because the uploads already carry the subject, the styling and the layout. A translate button converts a prompt written in another language to English before it is sent, which helps when the description is easier to write in your first language. Aspect ratios run from 1:1 to 21:9 plus auto, and the output is PNG by default with JPEG available.

Use this page when the image has to survive a close look: a shop listing, a printed menu, a hero banner that will be cropped for several placements. Use the standard Nano Banana page when you are still choosing between directions, since thirty credits a test lets you run a wider sweep. The decision is economic rather than technical. Both pages accept reference uploads and use native aspect ratios; one is tuned for iteration, the other for the final render, and the price difference is the whole point of keeping them separate.

The resolution selector changes the file, not the bill. Every run is charged the same whether you ask for 1K, 2K or 4K, so the choice is about where the image will be seen rather than what you can afford. Print, a full-width banner and a packaging mock need the largest size; a feed post, an avatar or a thumbnail does not. File size and download time are the only things that grow with the setting, which is why rendering everything at 4K and shrinking it later costs time without buying anything.

Treat the five upload slots as five jobs rather than five attempts. The first should carry the subject at its most recognizable angle, the second a view that settles what the back or the profile looks like, the third the material or surface you want kept, and the fourth the light you are matching. Keep the last for a layout or a pose. With the slots assigned, the prompt only has to describe what changes. Both modes cost the same, so deciding late whether a run needs references does not alter what you spend.

Cómo funciona Nano Banana Pro

1. 1

   Choose a mode

   Text to image starts from the prompt alone. Image to image opens the reference uploader, where you can add up to five images that anchor the subject, styling or layout of the result.

2. 2

   Write the change, not the whole scene

   In image to image, name the edit and the area it affects. Everything you leave out is carried by the references. If English is not your first language, press Translate and review the prompt before running it.

3. 3

   Set the frame and confirm the price

   Pick an aspect ratio from auto, 1:1, 3:4, 2:3, 4:5, 9:16, 16:9, 4:3, 3:2, 5:4 or 21:9, choose PNG or JPEG, then generate. The button shows the 240-credit charge first.

Para qué sirve Nano Banana Pro

• Final product images

  Move a direction you tested on the standard model into Pro for the listing photo, where edges, reflections and surface texture are inspected at full size.

• Campaign variants at native ratios

  Write one concept once, then render it at 1:1 for a feed and 9:16 for stories without cropping the composition afterwards.

• Characters and mascots

  Keep a face or costume consistent across a set by reusing the same references and changing only the scene and action in each prompt.

• Client presentations

  Produce the frames that go into a deck or a proposal, where a soft edge or a smeared detail is the difference between a yes and a revision request.

Nano Banana Pro de un vistazo

Model

Google Nano Banana Pro, run through KIE

Credit cost

240 per generation in both modes

Modes

Text to image, image to image

Reference images

Up to 5 per request, PNG/JPG/JPEG/WEBP, 10 MB each

Aspect ratios

Auto plus 1:1, 3:4, 2:3, 4:5, 9:16, 16:9, 4:3, 3:2, 5:4 and 21:9

Output

PNG by default, JPEG optional

Preguntas frecuentes sobre Nano Banana Pro

How much does a Nano Banana Pro image cost?

Each generation costs 240 credits, whether you use text to image or image to image, and whether you attach one reference or five. A free account starts with 300 bonus credits and can claim 50 more on each day you log in; those credits expire after 30 days, so the daily claim matters more than the balance.

Why does this page cost more than the Nano Banana page?

The standard page runs the base model at 30 credits and is built for testing directions. Pro costs 240 credits because it is the final-fidelity pass, and the page is meant for the frame you publish rather than the ten sketches before it. Both pages accept reference uploads and native aspect ratios.

How many reference images can I upload?

Up to five in one image-to-image request, as PNG, JPG, JPEG or WEBP files of no more than 10 MB each. Five is enough to hold a subject from several angles or combine a product with a background plate, and the reference set can be reused across prompts.

Does the price change with aspect ratio or output format?

No. The charge is 240 credits per run regardless of the ratio you pick, the PNG or JPEG output, or the number of references attached. The price is shown on the generate button before the request is sent, so there is no separate estimate to interpret.

What happens if a generation fails?

Credits are returned automatically when the provider does not deliver an image, including jobs that expire before finishing. You are charged for output rather than attempts, which is what makes it practical to try a difficult composition more than once.

Does the Translate switch let me read the English before the run starts?

No. Translation happens on the server as the request is submitted, and the server decides whether your prompt needs it at all, so the box is not rewritten in front of you. If the translation call fails, the run goes ahead with your original wording rather than stopping. Keep names, brand words and any text that has to appear inside the image in Latin characters; they are more likely to survive untranslated and to render the way you expect.